🚀 go-pugleaf

In message <4lgflf$c6h@clarknet.clark.net> - whitteng%/etc/HOSTNAME (Gary
Whitten)22 Apr 1996 17:29:19 GMT writes:
:>
:>I have just found out that there appears to be big problems with some
:>PCI/IDE controllers on, mostly, Intel motherboards. In particular, the
:>CMD640 or RZ1000. The problem occurs when communications are attempted
:>between devices on different IDE controllers. Small transfers are OK,
:>but large trasnfers eventually hang the sytem, requiring a
:>reboot. Very nasty. The Linux fix is to insert
:>
:>	hd*=serialize
:>at the LILO prompt. It seems to work.
:>
:>To avoid this problem in future systems, can anyone recommend a way to
:>make sure a new motherboard will NOT have these problems?
:>
:>	Thanks for your time,
:>		Gary Whitten
:>		whitteng@acoast.com


     The reason these chipsets are treated differently is because there's a
terrible bug associated with them...this bug can corrupt your data.  Most
OSs (including Linux) use workarounds to keep any data from getting
corrupted.  The earlier kernels need the LILO fix whereas later kernels have
it incorporated into the kernel.

     I don't know how anyone can protect themselves from this kind of thing
in the future, especially since companies like Intel won't come clean until
it's too late...at least the later 'Triton' MBs don't have this problem.

 ...Good Luck,

 ...Norm