I discovered recently that a DEC3000/300LX on which I have been running
NetBSD/alpha 1.4 without problems for a week or so actually fails one of
the extended memory tests.  (I was trying some RAM from an RS/6000 in
it, to see if it would work, which is why I happened to be using the
extended tests.)  The machine passes the boot-up memory test, but a
"test mem" from the PROM prompt eventually fails as follows:

>>> test mem
T-STS-MEM - LLSC Test: Addr 00800000 FWD Wr 00000000
? T-ERR-MEM - stl_c bcache miss with victim at Addr: 00d91bb8
T-STS-MEM - Uncorrected Error count = 1
? T-ERR-MEM - FAILED status = 20 test Init addr = 00d91bb8
?? 810       MEM 0x0020


  84 FAIL

>>>

The above output is actually from testing genuine DEC MS16-BA memory
upgrades (32MB in the form of four 8MB SIMMs) from another machine.  
The same SIMMs check out without errors in another DEC3000/300LX, so I
know *they* are not bad.

So, what is actually failing here?  Note: that this DEC3000/300LX also
fails the CXT test with what appears to be a stuck-at fault in the
framebuffer RAM.  Is there some other RAM that's failing in the LLSC
test other than the on-board SIMMs?  If so, is it on the CPU card, or
the mainboard?  I don't think the former, because I tried another CPU
card and got the same error.

What is the likely consequence of failing an LLSC test?  Like I said,
I've been running NetBSD without any (apparent) problems (I still am),
and just discovered the memory test failure by accident.  Is the LLSC
instruction used to implement semaphores?  If so, does this mean my
mutexes aren't really that?  However, my gut feeling tells me that my
system would've crashed long before now if concurrency control wasn't
working properly.

Can anyone shed any light on the ramifications of a failed LLSC memory
test?  Also, what can/should I do to try and fix it.

Apologies if this is more DEC3000/300 than NetBSD/alpha...

Cheers,

Paul.

e-mail: paul@gromit.dlib.vt.edu

"I don't live today; maybe tomorrow..."
	--- James Marshall Hendrix