Re: paging Oops in 2.4.1{4,5-pre1}

[Linus Torvalds]

In article <[e-mail]>,
Daniel I. Applebaum <[e-mail]> wrote:
>
>I have a problem with kernels 2.4.14 and 2.4.15-pre1.  As soon as my
>system needs to page, it generates the following Oops.  I can
>duplicate this error at will by running a few memory-intensive
>processes, such as 4-5 simultaneously compiles, StarOffice, and
>Netscape.  If I duplicate the test, but running with no swap, just
>RAM, then I get the expected "Out of Memory: Killed process..."
>errors.

It is jumping to la-la land, apparently from "do_swap_page()". The
interesting part there is that do_swap_page() doesn't even follow any
suspicious function pointers or anything..

Can you do a

	gdb vmlinux

and send me the output of "disassemble do_swap_page", along with a copy
of the oops (the latter just because I don't keep archives of
linux-kernel, so I don't want to have to search for the oops again).

Oh, and only the first oops tends to be the really interesting one -
after the kernel has oopsed once, kernel data structures are quite
possibly corrupt, and subsequent oopses are suspect. That do_swap_page
one _was_ the first oops, right?

		Linus