Re: (ITS#6745) HEAD emfile race condition -> slapd stopping listening?

[hyc@symas.com]

h.b.furuseth@usit.uio.no wrote:
> Full_Name: Hallvard B Furuseth
> Version: HEAD
> OS:
> URL:
> Submission from: (NULL) (129.240.6.233)
> Submitted by: hallvard
>
>
> In HEAD slapd/daemon.c, thanks to HEAD's multiple listener support:
>
>    slap_listener() protects emfile with slap_daemon[0 ].sd_mutex,
>    but slapd_remove() protects it  with slap_daemon[id].sd_mutex.
>
> Maybe to compensate, slapd_remove() has code which checks if emfile
> is too big, but nothing checks if it is too small - which looks like
> slapd might never start listening again.
>
> Simplest fix: add this to slapd_remove():
>
>     if (id) ldap_pvt_thread_mutex_<lock,unlock>(&slap_daemon[0].sd_mutex );
>
> slap_daemon[0].sd_mutex looks quite contended that way, though.
> Maybe a separate emfile_mutex is better.

Alternatively, just skip the check entirely when id != 0. Which would mean 
that only closing a session on the first listener would ever trigger the other 
listeners to be unmuted. Doesn't seem so terrible; if slapd is actually out of 
descriptors one or two connections either way won't make a huge difference in 
service.

-- 
   -- Howard Chu
   CTO, Symas Corp.           http://www.symas.com
   Director, Highland Sun     http://highlandsun.com/hyc/
   Chief Architect, OpenLDAP  http://www.openldap.org/project/