On Tue, Oct 25, 2005 at 20:42:47 +0400, Valeriy E. Ushakov wrote:

> I think I found the cache bug.  cache.c 1.79 changed cache flush
> routines to use precomputed number of cache lines instead of doing an
> extra access and shift each time.  The problem is that on split I/D
> cache systems c_nlines is not initialized.  Oops, no more cache
> flushes :)
> 
> I will commit a fix shortly and request pull-ups to 2.x and 3.x.  It
> is too late for 2.1, though, so pick it from the branch when it's
> pulled.

If you have a split I/D cache system and saw weird core dumps, please,
try cpu.c revision 1.193.

PS: This should probably fix "Bad pAUX_base" too.  Incidentally, the
first report about "Bad pAUX_base" was just three days after split
cache systems were broken.

SY, Uwe
-- 
uwe@ptc.spbu.ru                         |       Zu Grunde kommen
http://www.ptc.spbu.ru/~uwe/            |       Ist zu Grunde gehen