Damn, I thought I had this one nailed..
> It's unclear to me whether this crash is the same problem or a different one--
> the kernel oopses did look different, but unfortunately they weren't being
> logged to a file.
Could you please to try write them down by hand (yes, I know it is
extremely horrible to do) and write them back in once you reboot to run
ksymoops on them to get a symbolic trace?
The fact that you saw problems implies that your threads _are_ closing
files that are still in use by another thread, and it is possible that
"select()" was only one of the uses. The pre-89-3 kernel protects read(),
write() and select() (read/write have been there "forever"), but there are
other cases that it might not protect against (FASYNC comes to mind: can
you think of what operations your application would be using on the
sockets it has open in threads?).
This may explain why the test-program posted doesn't have problems (it
probably only does selects), but your load may have different patterns.
Linus
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu