[PATCH 3.2 055/221] libceph: fix double __remove_osd() problem

[Ben Hutchings]

3.2.69-rc1 review patch.  If anyone has any objections, please let me know.

------------------

From: Ilya Dryomov <idryomov@xxxxxxxxx>

commit 7eb71e0351fbb1b242ae70abb7bb17107fe2f792 upstream.

It turns out it's possible to get __remove_osd() called twice on the
same OSD.  That doesn't sit well with rb_erase() - depending on the
shape of the tree we can get a NULL dereference, a soft lockup or
a random crash at some point in the future as we end up touching freed
memory.  One scenario that I was able to reproduce is as follows:

            <osd3 is idle, on the osd lru list>
<con reset - osd3>
con_fault_finish()
  osd_reset()
                              <osdmap - osd3 down>
                              ceph_osdc_handle_map()
                                <takes map_sem>
                                kick_requests()
                                  <takes request_mutex>
                                  reset_changed_osds()
                                    __reset_osd()
                                      __remove_osd()
                                  <releases request_mutex>
                                <releases map_sem>
    <takes map_sem>
    <takes request_mutex>
    __kick_osd_requests()
      __reset_osd()
        __remove_osd() <-- !!!

A case can be made that osd refcounting is imperfect and reworking it
would be a proper resolution, but for now Sage and I decided to fix
this by adding a safe guard around __remove_osd().

Fixes: http://tracker.ceph.com/issues/8087

Cc: Sage Weil <sage@xxxxxxxxxx>
Signed-off-by: Ilya Dryomov <idryomov@xxxxxxxxx>
Reviewed-by: Sage Weil <sage@xxxxxxxxxx>
Reviewed-by: Alex Elder <elder@xxxxxxxxxx>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@xxxxxxxxxxxxxxx>
---
--- a/net/ceph/osd_client.c
+++ b/net/ceph/osd_client.c
@@ -678,12 +678,22 @@ static void put_osd(struct ceph_osd *osd
  */
 static void __remove_osd(struct ceph_osd_client *osdc, struct ceph_osd *osd)
 {
-	dout("__remove_osd %p\n", osd);
+	dout("%s %p osd%d\n", __func__, osd, osd->o_osd);
 	BUG_ON(!list_empty(&osd->o_requests));
-	rb_erase(&osd->o_node, &osdc->osds);
 	list_del_init(&osd->o_osd_lru);
-	ceph_con_close(&osd->o_con);
-	put_osd(osd);
+	rb_erase(&osd->o_node, &osdc->osds);
+	RB_CLEAR_NODE(&osd->o_node);
+}
+
+static void remove_osd(struct ceph_osd_client *osdc, struct ceph_osd *osd)
+{
+	dout("%s %p osd%d\n", __func__, osd, osd->o_osd);
+
+	if (!RB_EMPTY_NODE(&osd->o_node)) {
+		ceph_con_close(&osd->o_con);
+		__remove_osd(osdc, osd);
+		put_osd(osd);
+	}
 }
 
 static void remove_all_osds(struct ceph_osd_client *osdc)
@@ -693,7 +703,7 @@ static void remove_all_osds(struct ceph_
 	while (!RB_EMPTY_ROOT(&osdc->osds)) {
 		struct ceph_osd *osd = rb_entry(rb_first(&osdc->osds),
 						struct ceph_osd, o_node);
-		__remove_osd(osdc, osd);
+		remove_osd(osdc, osd);
 	}
 	mutex_unlock(&osdc->request_mutex);
 }
@@ -723,7 +733,7 @@ static void remove_old_osds(struct ceph_
 	list_for_each_entry_safe(osd, nosd, &osdc->osd_lru, o_osd_lru) {
 		if (time_before(jiffies, osd->lru_ttl))
 			break;
-		__remove_osd(osdc, osd);
+		remove_osd(osdc, osd);
 	}
 	mutex_unlock(&osdc->request_mutex);
 }
@@ -739,7 +749,7 @@ static int __reset_osd(struct ceph_osd_c
 	dout("__reset_osd %p osd%d\n", osd, osd->o_osd);
 	if (list_empty(&osd->o_requests) &&
 	    list_empty(&osd->o_linger_requests)) {
-		__remove_osd(osdc, osd);
+		remove_osd(osdc, osd);
 	} else if (memcmp(&osdc->osdmap->osd_addr[osd->o_osd],
 			  &osd->o_con.peer_addr,
 			  sizeof(osd->o_con.peer_addr)) == 0 &&
@@ -1271,6 +1281,7 @@ static void reset_changed_osds(struct ce
 {
 	struct rb_node *p, *n;
 
+	dout("%s %p\n", __func__, osdc);
 	for (p = rb_first(&osdc->osds); p; p = n) {
 		struct ceph_osd *osd = rb_entry(p, struct ceph_osd, o_node);
 

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/