ec/fini: Fix race with ec_fini and ec_notify

During a graph cleanup, we first sent a PARENT_DOWN and wait for
a child down to ultimately free the xlator and the graph.

In the ec xlator, we cleanup the threads when we get a PARENT_DOWN event.
But a racing event like CHILD_UP or event xl_op may trigger healing threads
after threads cleanup.

So there is a chance that the threads might access a freed private variabe

Change-Id: I252d10181bb67b95900c903d479de707a8489532
fixes: bz#1703948
Signed-off-by: Mohammed Rafi KC <rkavunga@redhat.com>

6 files changed, 38 insertions, 0 deletions

diff --git a/libglusterfs/src/glusterfs/xlator.h b/libglusterfs/src/glusterfs/xlator.h
index b7d0249c9c8..1a0ed99774b 100644
--- a/libglusterfs/src/glusterfs/xlator.h
+++ b/libglusterfs/src/glusterfs/xlator.h
@@ -1092,4 +1092,7 @@ gluster_graph_take_reference(xlator_t *tree);
 
 gf_boolean_t
 mgmt_is_multiplexed_daemon(char *name);
+
+gf_boolean_t
+xlator_is_cleanup_starting(xlator_t *this);
 #endif /* _XLATOR_H */
diff --git a/libglusterfs/src/libglusterfs.sym b/libglusterfs/src/libglusterfs.sym
index 1955005a88f..93ab05f166a 100644
--- a/libglusterfs/src/libglusterfs.sym
+++ b/libglusterfs/src/libglusterfs.sym
@@ -1164,3 +1164,4 @@ glusterfs_process_svc_attach_volfp
 glusterfs_mux_volfile_reconfigure
 glusterfs_process_svc_detach
 mgmt_is_multiplexed_daemon
+xlator_is_cleanup_starting
diff --git a/libglusterfs/src/xlator.c b/libglusterfs/src/xlator.c
index 292416ae449..526cd2e2e4d 100644
--- a/libglusterfs/src/xlator.c
+++ b/libglusterfs/src/xlator.c
@@ -1508,3 +1508,24 @@ mgmt_is_multiplexed_daemon(char *name)
     }
     return _gf_false;
 }
+
+gf_boolean_t
+xlator_is_cleanup_starting(xlator_t *this)
+{
+    gf_boolean_t cleanup = _gf_false;
+    glusterfs_graph_t *graph = NULL;
+    xlator_t *xl = NULL;
+
+    if (!this)
+        goto out;
+    graph = this->graph;
+
+    if (!graph)
+        goto out;
+
+    xl = graph->first;
+    if (xl && xl->cleanup_starting)
+        cleanup = _gf_true;
+out:
+    return cleanup;
+}
diff --git a/xlators/cluster/ec/src/ec-heal.c b/xlators/cluster/ec/src/ec-heal.c
index 2fa1f11c5fa..8844c292f83 100644
--- a/xlators/cluster/ec/src/ec-heal.c
+++ b/xlators/cluster/ec/src/ec-heal.c
@@ -2855,6 +2855,10 @@ ec_replace_brick_heal_wrap(void *opaque)
         itable = ec->xl->itable;
     else
         goto out;
+
+    if (xlator_is_cleanup_starting(ec->xl))
+        goto out;
+
     ret = ec_replace_heal(ec, itable->root);
 out:
     return ret;
diff --git a/xlators/cluster/ec/src/ec-heald.c b/xlators/cluster/ec/src/ec-heald.c
index 985d4854363..5948a588be1 100644
--- a/xlators/cluster/ec/src/ec-heald.c
+++ b/xlators/cluster/ec/src/ec-heald.c
@@ -444,6 +444,9 @@ unlock:
 int
 ec_shd_full_healer_spawn(xlator_t *this, int subvol)
 {
+    if (xlator_is_cleanup_starting(this))
+        return -1;
+
     return ec_shd_healer_spawn(this, NTH_FULL_HEALER(this, subvol),
                                ec_shd_full_healer);
 }
@@ -451,6 +454,9 @@ ec_shd_full_healer_spawn(xlator_t *this, int subvol)
 int
 ec_shd_index_healer_spawn(xlator_t *this, int subvol)
 {
+    if (xlator_is_cleanup_starting(this))
+        return -1;
+
     return ec_shd_healer_spawn(this, NTH_INDEX_HEALER(this, subvol),
                                ec_shd_index_healer);
 }
diff --git a/xlators/cluster/ec/src/ec.c b/xlators/cluster/ec/src/ec.c
index 4e9fac5ea39..95100951220 100644
--- a/xlators/cluster/ec/src/ec.c
+++ b/xlators/cluster/ec/src/ec.c
@@ -486,6 +486,9 @@ ec_set_up_state(ec_t *ec, uintptr_t index_mask, uintptr_t new_state)
 {
     uintptr_t current_state = 0;
 
+    if (xlator_is_cleanup_starting(ec->xl))
+        return _gf_false;
+
     if ((ec->xl_notify & index_mask) == 0) {
         ec->xl_notify |= index_mask;
         ec->xl_notify_count++;