Re: [sshfs] SSHFS gets stuck if connection is lost

[Miklos S. <mi...@sz...>]

On Tue, 25 Jan 2011, Thue Janus Kristensen wrote:
> I just tried starting both sshfs and a normal openssh sftp client at the
> same time, both against the same server, as an experiment.
> 
> When I rebooted the remote server, the ssh client said "Connection to
> thuejk.dk closed by remote host.". The sshfs mount didn't say anything, and
> will still hang any "ls" command which tries to read from the sshfs mount
> point.
> 
> At least in this case, it seems that sshfs should know that the other end is
> going away permanently, but fails to act appropriately.

Yeah, it should, but it doesn't.  Following patch fixes it for me.

Thanks,
Miklos


diff --git a/sshfs.c b/sshfs.c
index 61f2b92..81ee64a 100644
--- a/sshfs.c
+++ b/sshfs.c
@@ -1314,17 +1314,16 @@ static void *process_requests(void *data_)
 			break;
 	}
 
+	pthread_mutex_lock(&sshfs.lock);
+	sshfs.processing_thread_started = 0;
+	close_conn();
+	g_hash_table_foreach_remove(sshfs.reqtab, (GHRFunc) clean_req, NULL);
+	sshfs.connver ++;
+	pthread_mutex_unlock(&sshfs.lock);
+
 	if (!sshfs.reconnect) {
 		/* harakiri */
 		kill(getpid(), SIGTERM);
-	} else {
-		pthread_mutex_lock(&sshfs.lock);
-		sshfs.processing_thread_started = 0;
-		close_conn();
-		g_hash_table_foreach_remove(sshfs.reqtab, (GHRFunc) clean_req,
-					    NULL);
-		sshfs.connver ++;
-		pthread_mutex_unlock(&sshfs.lock);
 	}
 	return NULL;
 }