==> Regarding [Wlug] System hang woes continue; Andy Stewart <andystewart@comcast.net> adds: andystewart> -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 andystewart> HI gang, andystewart> I think I somehow built the computer from Hell. Recall my andystewart> dual Opteron box with SuSE 9.2 and 1 GB of memory. andystewart> I removed every SATA device from my system, and now it doesn't andystewart> hang as often, but it still hangs periodically. This time I andystewart> got 10 days of uptime before it hung - a new world's record on andystewart> this particular box. Strangely enough, I could demonstrate andystewart> more uptime with the "old" SATA code than with libata - go andystewart> figure. andystewart> I am suspicious of the 2.6 kernel. We run dual Opteron andystewart> servers at work with the 2.4 kernel series with no problems at andystewart> all (on RedHat 7.3). I am wondering what would happen if I andystewart> took SuSE 9.2 and replaced the 2.6 kernel with a 2.4 kernel. andystewart> Am I asking for a heap of trouble? andystewart> What specific issues would I encounter? Problems with your device tree, no doubt. I think this is a rat hole. Just don't go there. I'd run a live cd based on 2.4 before trying what you're suggesting. andystewart> I'm guessing that the kernel modules would be all fubar since andystewart> I thought taht 2.4 and 2.6 did that quite differently. andystewart> Perhaps I could bypass that with a monolithic kernel. I'm not andystewart> sure what other problems I'd make for myself if I did this. andystewart> Another thought - does somebody have a .config file for a andystewart> Linux 2.6 kernel on an Opteron system that works really well? The distribution configs always work for me. They probably see the most testing, if you think about it. One other suggestion: have you checked to see if there are any BIOS updates? I've seen *very* strange problems of the random variety that are really BIOS bugs. If you're really serious about tracking this down, I'd start doing some stress testing to see if you can get a reliable reproducer. Without that, it's tough to address. Or, enable kexec/kdump. There's a quick blurb on it in fedora weekly news, but that's for fedora. -Jeff