Messages in this thread | | | Date | Tue, 15 Feb 2000 19:03:22 -0500 | From | "Theodore Y. Ts'o" <> | Subject | Re: HARDWARE HELL - POSSIBLE LINUX BUG AND/OR HARDWARE FAILURE ? |
| |
Something that you might want to do is to check the temperature of various parts of your system on the motherboard. There have been cases where memory that's been certified by a major CPU manufacturer as being Good Stuff was, when run under a really good memory torture test (written by Larry Augustine), was observed to reach surface temperatures of over 90 degrees centigrade. (And this was in an otherwise well ventilated chasis!) The guess is that it worked fine in Windows sytems, because it never agressively accessed the memory that hard, so the manufacturer never noticed the fundamental design flaw. Obviously, we decided not to use that particular memory in our systems. :-)
Given that your problems started (and restarted) when you ran GNU shred, I would suggest running it, and then when your system crashes, take some temperature measurements and see if anything seems unusually hot. From your description, it sounds suspiciously like a thermal problem. Do the problems seem to go away (or lessen) if you let the computer cool down for a while?
If you have some temperature probes, try attaching them to various parts of the motherboard (especially the CPU, memory, and cache), and then run your system with the cover on. The problem may also be exacerbated by a bad chasis design so that there are "hot spots" in certain parts of your motherboard. Then, when a program agressively exercises the system, it's enough to cause some component (probably memory or cache, given your symptoms) to fall over. Oh, I'd also put a thermal sensor on your disk drive; some disk drives can get *really* hot when you exercise them a lot. (The Seagate Barracudas come to mind as a good example.)
Finally, I'll give a plug for VA Linux's "burn-in" software, which can be found at "ftp://ftp.valinux.com/pub/software/Cerberos". You might find it a good way to torture test a system (sometimes to failure, if it's dodgy on ventilation, power, etc.; read the warnings before you use it.)
- Ted
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |