On Tue, Jun 10, 2025 at 08:41:44AM -0700, Gleb Smirnoff wrote:
> On Tue, Jun 10, 2025 at 12:11:18PM +0300, Konstantin Belousov wrote:
> K> Apparently we already have the feature called 'warmstart', where rpcbind
> K> can be restarted and existing registrations are reloaded. So it is in
> K> fact already solved, assuming admin is careful enough to use -w.
>
> Ok, then all that needs to be done is revert this:
>
> --- a/usr.sbin/rpcbind/rpcbind.c
> +++ b/usr.sbin/rpcbind/rpcbind.c
> @@ -159,10 +159,6 @@ main(int argc, char *argv[])
>
> update_bound_sa();
>
> - /* Ensure krpc is loaded */
> - if (modfind("krpc") < 0 && kldload("krpc") < 0)
> - err(1, "krpc");
> -
> /* Check that another rpcbind isn't already running. */
> if ((rpcbindlockfd = open(RPCBINDDLOCK, O_RDONLY|O_CREAT, 0444)) < 0)
> err(1, "%s", RPCBINDDLOCK);
>
> Without these lines, rpcbind will log:
>
> Jun 10 08:34:23 bobrik rpcbind[885]: netlink: could not create service
Well, absolutely minimal fix would be to change err() to warn().
>
> And continue with /var/run/rpcbind.sock set up and ready to service
> requests from applications.
>
> We can add an extra message for verbosity:
>
> @@ -223,7 +219,10 @@ main(int argc, char *argv[])
> }
> endnetconfig(nc_handle);
>
> - init_transport(&netlink_nconf);
> + if (init_transport(&netlink_nconf) != 0) {
> + syslog(LOG_ERR, "krpc(4) module missing? "
> + "Continuing without netlink/krpc transport");
> + }
>
> /*
> * Allocate pipe fd to wake main thread from signal handler in
> non-racy
>
> I really don't have a strong opinion on what would be right here. I have no
> idea on how wide is the use of rpcbind w/o NFS. Kostik, Rick and Dag-Erling,
> may you together come to agreement on what is right here?
>
> --
> Gleb Smirnoff
