What do the Winbind 2.2.2 memory leaks mean?

Michael Steffens michael_steffens at hp.com
Sun Dec 16 05:12:02 GMT 2001

Hi Alexander,

Alexander Bokovoy wrote:
> > Why doesn't just restarting Winbind and/or Samba do the job?
> > What kind of memory is being leaked there?
> Samba 2.2.3 CVS (SAMBA_2_2 branch) has these leaks fixed.

yes I know, and I appreciate that! But it doesn't answer my
question, which is mostly for curiosity.

The kind of memory leak I'm used to causes a process to
malloc more and more memory, failing to free it again,
until there is nothing left to be malloc'ed any more.

However, if the process is killed and restarted, all
resources are being released and the game can start from
the beginning. If leaking is slow and regular restarts
are feasible, this wouldn't even be that much of a problem.

But what I observed with 2.2.2 Winbind was different. It was
restarted now and than in order to recover from hangs, but
after a while of doings so (approximately one month) restarting
didn't help any more. But a reboot of the entire machine did.

Thus my question: what kind of memory is being leaked there,
such that is persists process restart, but not reboot? Is
it shared memory?

Did others observe this misbehaviour, too? Do the fixes in
2.2.3 apply to that, or to the "classic" kind of leak, which
should be recoverable by a simple process restart.


More information about the samba-technical mailing list