[stgt] help tgt segfault

[FUJITA Tomonori]

On Wed, 17 Dec 2008 16:09:58 +0100
Tomasz Chmielewski <mangoo at wpkg.org> wrote:

> Tomasz Chmielewski schrieb:
> > FUJITA Tomonori schrieb:
> > 
> >>> Which will result in:
> >>>
> >>> Dec 16 13:55:32 megathecus kernel: tgtd[4872]: segfault at 00000220 
> >>> eip 0804f0b5 esp 77c43ff0 error 4
> >>
> >> Hmm, unfortunately, I can't reproduce this.
> > 
> > Strange, it is 100% reproducible here, on a x86 system. I didn't try 64 
> > bit.
> > I just do dmsetup suspend / wait a few seconds / dmsetup resume, while 
> > the initiator is reading data, and tgtd segfaults.
> 
> It crashes if I wait about 120 seconds or more.
> If I wait less, it does not crash.

I made sure that I wait about 120 seconds but I can't reproduce this
(it takes about 120 seconds to see the abort and close messages in the
log.

We should not see a series of 'conn_close' messages. Something
unexpected happens at your box.

Can you try this and send the log? It doesn't fix the problem but
gives more information.

diff --git a/usr/iscsi/conn.c b/usr/iscsi/conn.c
index c205397..7a626c9 100644
--- a/usr/iscsi/conn.c
+++ b/usr/iscsi/conn.c
@@ -82,10 +82,12 @@ void conn_exit(struct iscsi_connection *conn)
 void conn_close(struct iscsi_connection *conn)
 {
 	struct iscsi_task *task, *tmp;
+	int ret;
 
-	conn->tp->ep_close(conn);
+	ret = conn->tp->ep_close(conn);
 
-	eprintf("connection closed %p %u\n", conn, conn->refcount);
+	eprintf("connection closed %p %u %d %d\n",
+		conn, conn->refcount, ret, errno);
 
 	/* may not have been in FFP yet */
 	if (!conn->session)
diff --git a/usr/iscsi/iscsi_tcp.c b/usr/iscsi/iscsi_tcp.c
index 2320b3e..0007e66 100644
--- a/usr/iscsi/iscsi_tcp.c
+++ b/usr/iscsi/iscsi_tcp.c
@@ -164,8 +164,8 @@ static void iscsi_tcp_event_handler(int fd, int events, void *data)
 		iscsi_tx_handler(conn);
 
 	if (conn->state == STATE_CLOSE) {
+		eprintf("connection closed %p\n", conn);
 		conn_close(conn);
-		dprintf("connection closed\n");
 	}
 }
 
diff --git a/usr/iscsi/iscsid.c b/usr/iscsi/iscsid.c
index c22a6f6..479dcfa 100644
--- a/usr/iscsi/iscsid.c
+++ b/usr/iscsi/iscsid.c
@@ -1085,6 +1085,8 @@ static int iscsi_scsi_cmd_done(uint64_t nid, int result, struct scsi_cmd *scmd)
 	 * task got reassinged to another connection.
 	 */
 	if (task->conn->state == STATE_CLOSE) {
+		eprintf("connection already closed %p %u\n",
+			task->conn, task->conn->refcount);
 		iscsi_free_cmd_task(task);
 		return 0;
 	}
diff --git a/usr/tgtd.c b/usr/tgtd.c
index 758e7d5..f29dab1 100644
--- a/usr/tgtd.c
+++ b/usr/tgtd.c
@@ -137,6 +137,7 @@ static struct event_data *tgt_event_lookup(int fd)
 void tgt_event_del(int fd)
 {
 	struct event_data *tev;
+	int ret;
 
 	tev = tgt_event_lookup(fd);
 	if (!tev) {
@@ -144,7 +145,10 @@ void tgt_event_del(int fd)
 		return;
 	}
 
-	epoll_ctl(ep_fd, EPOLL_CTL_DEL, fd, NULL);
+	ret = epoll_ctl(ep_fd, EPOLL_CTL_DEL, fd, NULL);
+	if (ret < 0)
+		eprintf("fail to remove epoll event %d %d\n", ret, errno);
+
 	list_del(&tev->e_list);
 	free(tev);
 }
--
To unsubscribe from this list: send the line "unsubscribe stgt" in
the body of a message to majordomo at vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html