Sourceforge 1.5pre3 (and earlier) RPMS have broken flock support on Linux


Subject: Sourceforge 1.5pre3 (and earlier) RPMS have broken flock support on Linux
From: Kevin M. Myer (kevin_myer@iu13.k12.pa.us)
Date: Fri Jan 12 2001 - 14:49:38 EST


Hello,

I posted a message a few weeks ago about a problem I was seeing here
relating to file locking. I was running a stock netatalk 1.5pre3
installation from the Sourceforge RPM and it was brought to my attention
by our users that there could be multiple read/write instances for any
file on the server. Upon investigation, I found that here was not a
single lock being created. I received no responses from the lists so I'm
not sure if I'm an isolated case or if no one cares that their netatalk
servers aren't locking :)

Anyway, after a lot of testing and trying to understand how the locking
was working and a lot of recompiling, I figured out that the
--enable-flock-locks option to configure was what was breaking the
locking. When I just let if use fcntl to do the locking, everything works
perfectly.

What this means is that at the very least, the Sourceforge 1.5pre3 (and
probably earlier releases) is broken with respect to x86 Linux
architectures. I think I recall reading somewhere that flock was at one
time broken under Linux so maybe its an operating system issue. In any
case, I'd recommend against using --enable-flock-locks if you're using x86
Linux.

So far, I've reproduced the problem on every x86 machine I've tried. I
thought it might be related to the kernel on our SMP machine which has a
large file system patch on it but a uniprocessor stock 2.2.16 kernel has
the some locking problems.

If anyone could enlighten me as to whether this is a kernel issue or if
its netatalk specific, it would help with further isolating the bug.

Thanks,

Kevin

-- 
Kevin M. Myer
Systems Administrator
Lancaster-Lebanon Intermediate Unit 13
(717)-560-6140



This archive was generated by hypermail 2b28 : Sun Oct 14 2001 - 03:04:30 EDT