[SSSD] [PATCH] DP: Notify propperly when removing PAC responder

Jakub Hrozek jhrozek at redhat.com
Mon Aug 19 19:20:10 UTC 2013


On Thu, Aug 08, 2013 at 01:06:24AM +0200, Pavel Březina wrote:
> On 08/08/2013 12:37 AM, Jakub Hrozek wrote:
> >On Thu, Aug 08, 2013 at 12:32:37AM +0200, Pavel Březina wrote:
> >>On 08/05/2013 04:36 PM, Ondrej Kos wrote:
> >>>Hi,
> >>>
> >>>During testing, I noticed, that all responders are logged propperly when
> >>>being shut down:
> >>>
> >>>[sssd[be[DOM]]] [be_client_destructor] (0x0400): Removed PAM client
> >>>[sssd[be[DOM]]] [be_client_destructor] (0x0400): Removed NSS client
> >>>
> >>>But not PAC:
> >>>
> >>>[sssd[be[DOM]]] [be_client_destructor] (0x0020): Unknown client removed ...
> >>>
> >>>Attached patch adds pac_cli pointer, to fix this behavior.
> >>>
> >>>Ondra
> >>
> >>Ack.
> >
> >Do the _cli pointers actually serve any purpose except for the
> >backchannel towards nss responder? Shouldn't we rather clean them up?
> 
> That's a good question.
> 
> Only nss_cli is used, but in initgroups callback where it can be
> accessed by be_req->becli (which is used by other callbacks).

Sumit, do you remember why we put the explicit comment to
client_registration() ? Was it just because we didn't plan on using the
pointer explicitly? This one:

@@ -2118,7 +2121,7 @@ static int client_registration(DBusMessage *message,
     } else if (strcasecmp(cli_name, "SSH") == 0) {
         becli->bectx->ssh_cli = becli;
     } else if (strcasecmp(cli_name, "PAC") == 0) {
-        /* no need to set becli */
+        becli->bectx->pac_cli = becli;
     } else {
         DEBUG(1, ("Unknown client! [%s]\n", cli_name));
     }


I think we should commit this patch to get rid of the warning
and then file another ticket that we would solve together with
https://fedorahosted.org/sssd/ticket/2038 to review if we actually need
all these client pointers..



More information about the sssd-devel mailing list