On Mon, 10 Jan 2005 19:41:01 -0800, Chris Lynch <tsw@xxxxxxx> wrote: > Sorry about the HTML msg and appreciate the help :S No problem. I don't really care all that much, but there are valid concerns about bandwidth for some people, others point out potential security risks. Also, it's a sort of standard around here to reply to a message below (or interspersed) and not above (called "top posting"), with the previous message quoted (ie indented with '>' ) and trimmed. These are things I had not heard of when I joined, so I try to point them out when I help. Anyway. > I tried the noapic option on boot and I guess it's working, but I don't see > a way to visually confirm other than I typed it in. I then tried ncftp and > got a segmentation fault (output appended at the bottom of this email). I > then tried to do a wget to see if port 80 was any different and also got a > segmentation fault. First of all, the things you have been posting are not segmentation faults (a program tries to access memory it doesn't have), but kernel "Oops"es. A seg fault occurs in user land, a kernel Oops means the kernel ran into a problem (and dereferencing a NULL pointer is a *problem*). Seg faults can be fairly common and related to bad programs. The kernel, however, is should be much better, and Oopses should be rare if they happen at all. (This is all theoretical, of course : ) ) Given that this happens with a lot of network traffic it does seem to indicate an interrupt problem. If you did disable APIC and it didn't help, then I'm not really sure what is going on. You could also try disabling ACPI (yes, it is different : ) ) with "acpi=off" to see if that calms it down. Basically, it seems the computer is doing something fancy (with interrupts?), and then getting confused. Perhaps some more information about your system might help. Hardware info (network card, CPU, RAM, motherboard). How long has this been happening? Has it ever worked reliably? Do you have other OSs with which to test it? Perhaps you could try a bootable CD (knoppix, etc) at least to see if it is Fedora specific. Perhaps try running Memtest-86 for a while to be sure components are good. If you could do the same thing several times in a row and compare the Ooops output, that might help. Maybe we could find a similarity, or no pattern at all indicates something rather complex, like interrupts again. > The odd thing is that both only happen after close to a minute of > transferring. They both work fine for a few hundred megs, but then they > die. > > Any ideas? Is there any way to confirm that my noapic option is "working"? Check the output of the "demsg" command (from your post, you are familiar with it already). To actually see everything you can do: dmesg | more or for a more succinct version: dmesg | grep -i apic Look for stuff about noapic and disabling PIC. Also, compare your /proc/interrupts to before, you should see some differences. > Chris > [snip] Jonathan
