pyxis_machine_check - what on Earth is this?!?!?!

[abel bfr co il (Alexander L. Belikoff)]

Hello everybody,

I'm still looking for a stable kernel(tm) for:

- LX164
- SRM
- serial console
- DEC Tulip
- RedHat 4.2

So far, my search was fruitless:

1. Original kernel - heavily patched 2.0.30 w/ a tulip.c 0.83. This
   one has a lot of memory problems. After about 40 days of uptime and 
   high network activity it hangs and starts spitting messages like:

	Memory squeeze... Deferring packet...

   BTW, when locking up it renders the TCP stack to such a screwed up
   state that an attempt to connect to the machine via [rs]sh makes
   the latter hang and stay like that with no timeout for weeks(!!!).


2. 2.0.35-0.2 w/ a de4x5.c driver (the one supplied with the kernel).
   This one has two problems:

   - Upon bootup, the net interface seems to be dead and it requires a 
     couple ping packets to revive it befor it starts working.

   - (the subject of this pathetic letter) Recently, I've observed two 
     network outages with this kernel. It happened on machines with
     relatively low uptimes (6 days and 1 day), whereas a number of
     other hosts have been running this kernel with no problems for
     more than 30 days. In both cases the net interface went down with 
     the following messages:

--------------------------------------------------------------------
PYXIS machine check NOT expected
pyxis_machine_check: vector=0x670 pc=0xfffffc00003ddf64
 +0 260 1a000000118
 +10 98 0
 +20 ffffff000001d02f 0
 +30 7 ffffffffffffffff
 +40 ffffff000042b2cf 6068
 +50 428e10 ffffffffffffffff
 +60 fffffc8580000000 fffffc00003115e8
 +70 5200 fffffc000043b860
 +80 11 0
 +90 fffffc0000311328 1f1e161514020100
 +a0 fffffc00003117f8 fffffc00003ddf64
 +b0 fffffc00003111a0 fffffc00003112b0
 +c0 100000000 c6008051
 +d0 7f457c2 9806700001
 +e0 0 0
 +f0 fffffc0000303fb0 300000
 +100 fffffc0000311238 fffffc000044ad98
 +110 428e10 fffffc00003ddf64
 +120 0 0
 +130 18000 0
 +140 4164020000 0
 +150 0 155558275d8
 +160 14ed0 ffffff000001d02f
 +170 0 ffffff80072f6fff
 +180 ffffffffffffffff 130
 +190 fffffff005ffffff ffffff000042b2cf
 +1a0 ffffffff80000080 10
 +1b0 b9b b00410000fb20
 +1c0 c0c 6385b0
 +1d0 58000000 0
 +1e0 2050000 48828
 +1f0 48828 200
 +200 43 c0
 +210 ff e9
 +220 ff 0
 +230 0 0
 +240 0 20
 +250 bf ef
eth0: Fatal bus error occurred, sts=0xffffffff, device stopped.
-------------------------------------------------------------------------


3. 2.0.35-0.2 but with a tulip.c - haven't tried it yet.


Therefore the questions:

1. Please please please: what the heck is this pyxis thingie and how
   can I fix it? Looks like all good guys have moved to 2.1+ kernels,
   yet I cannot afford such a luxury being stuck with RedHat 4.2
   (glibc 1.9)

2. Has anybody experienced similar problems with the de4x5 driver? I'd 
   like to hear suggestions about it.

3. Same for the tulip.c - which one of a couple dozens of these ones
   hanging around is stable enough to use?

Thanks in advance,


-- 
Alexander L. Belikoff
Bloomberg L.P. / BFM Financial Research Ltd.
abel@vallinor4.com, abel@bfr.co.il