Re: [wtjones@widowmaker.com] libc/836: There appears to be a bug in the sigmask handling in Linuxthreads.

[Thorsten Kukuk <kukuk@weber-eb.uni-paderborn.de>]

Hello,

> 
> Here's another bug report which I don't know how to handle.  The
> information is to vague to reproduce it - but has anybody an idea what 
> might be the problem?

Yes, I have.

> Uli, you write in the FAQ:
> } 2.23.   When I use nscd the machine freezes.
> } 
> } {UD} It is well known that you cannot use nscd with Linux 2.0.*.  There
> } is functionality missing in the kernel and work-arounds are not suitable.
> } Beside this some parts of the kernel are too buggy when it comes to using
> } threads.
> } 
> } So you have the possibilities to run Linux 2.0.* or update to a higher
> } version and start using nscd.
> 
> Might it be that the bug is a bug of Linux 2.0.35 and not of
> glibc/linuxthreads?

When developing the first nscd version, I found a bug in the linux
kernel: if one of the file handles, which select in waiting for, is
used in another thread, the program or the kernel or the system could
hang. This bug was known to all kernel developer. The most problems
with this should be fixed in the current development kernel, but I'm
not sure if all problems are fixed.

nscd uses poll, and caling the poll emulation for 2.0 kernels from
nscd causes the kernel to freeze. But since other code using poll
doesn't make problems, I think this is a kernel/select/thread problem.

If I look at the debug output from mysqld, it seems to make the same
things:
mysqld has one thread for select, which spawns new working threads.
Somebody should test, if mysqld works with linux 2.1.125. If it works,
it is the known kernel bug. If not, it is a glibc/linuxthreads bug.

  Thorsten

-- 
Thorsten Kukuk  kukuk@vt.uni-paderborn.de
                http://home.pages.de/~kukuk/
Linux is like a Vorlon.  It is incredibly powerful, gives terse,
cryptic answers and has a lot of things going on in the background.