[Discuss] Linux on Lenovo P70 -- data corruption

[rlk at alum.mit.edu (Robert Krawitz)]

I'm about at my wit's end here.

I bought a Lenovo P70 on eBay to run Linux on.  But I'm finding some
data corruption on network data transfers (e. g. scp, ftp, socat).
The oddity is that in the received files the bytes that are bad always
have 011111 in the low 6 bits of their address.  There's some
clustering, but not tight clustering and not at regular intervals.

rsync sometimes reports protocol failures (in either direction -- from
this machine to something else or from another system to this).  The
other system is good.  scp/ssh don't report any such problems.

It appears that this is more severe under heavy traffic, and perhaps
under heavy load too.

Unfortunately, I'm not able (yet) to reproduce this under Windows
using either ftp or scp, with various background loads.  The addresses
in suggest that something's going wrong with writing the data out to
disk (particularly with scp, I'd expect errors to propagate with
encryption or to get outright protocol failures), but the protocol
failures with rsync suggest that it's somewhere else.

I get no problems with either memtest86 or Lenovo diags, and prime95
also isn't reporting any issues.  This happens under both Knoppix
7.7.1 and openSUSE 42.3, the latter with any of the stock kernel, the
4.12.9+SUSE kernel, or the 4.12.9 vanilla kernel.

If I can reproduce the problem under Windows, the seller will refund
me in full and send it back to Lenovo.  If I can't, at best I'll have
to pay a 15% restocking fee on a not-inexpensive system.  If I try to
send it to Lenovo under warranty, I expect I'll get an NPF and just
waste a lot of time and yet another shipping fee.

The system has a Xeon E3-1505Mv5, nVidia M4000M (doesn't matter if X
is running or not), 32 GB RAM (yes, I've tried each of the two DIMMs
separately, in different slots), 4K display, two separate eSATA M.2
SSD's.  The BIOS is up to date with the hyperthreading fix, and in any
event it happens even if I turn hyperthreading off.  Right now it's an
expensive white elephant.

Any thoughts?
-- 
Robert Krawitz                                     <rlk at alum.mit.edu>

***  MIT Engineers   A Proud Tradition   http://mitathletics.com  ***
Member of the League for Programming Freedom  --  http://ProgFree.org
Project lead for Gutenprint   --    http://gimp-print.sourceforge.net

"Linux doesn't dictate how I work, I dictate how Linux works."
--Eric Crampton