afr : get split-brain-status in a synctask

        Backport of: http://review.gluster.org/#/c/12163/

On executing `getfattr -n replica.split-brain-status <file>` on mount,
there is a possibility that the mount hangs. To avoid this hang,
fetch the split-brain-status of a file in synctask.

 >Change-Id: I87b781419ffc63248f915325b845e3233143d385
 >BUG: 1262345
 >Signed-off-by: Anuradha Talur <atalur@redhat.com>

Change-Id: I9f4f4b54e108d3a0017264353b8272e072170c16
BUG: 1262547
Signed-off-by: Anuradha Talur <atalur@redhat.com>
Reviewed-on: http://review.gluster.org/12166
Tested-by: NetBSD Build System <jenkins@build.gluster.org>
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Pranith Kumar Karampuri <pkarampu@redhat.com>

6 files changed, 103 insertions, 22 deletions

diff --git a/xlators/cluster/afr/src/afr-common.c b/xlators/cluster/afr/src/afr-common.c
index 2eb3946ab7c..1c652f41e88 100644
--- a/xlators/cluster/afr/src/afr-common.c
+++ b/xlators/cluster/afr/src/afr-common.c
@@ -4588,8 +4588,7 @@ out:
 }
 
 int
-afr_get_heal_info (call_frame_t *frame, xlator_t *this, loc_t *loc,
-                   dict_t *xdata)
+afr_get_heal_info (call_frame_t *frame, xlator_t *this, loc_t *loc)
 {
         gf_boolean_t    data_selfheal     = _gf_false;
         gf_boolean_t    metadata_selfheal = _gf_false;
@@ -4714,7 +4713,14 @@ out:
 }
 
 int
-afr_get_split_brain_status (call_frame_t *frame, xlator_t *this, loc_t *loc)
+afr_get_split_brain_status_cbk (int ret, call_frame_t *frame, void *opaque)
+{
+        GF_FREE (opaque);
+        return 0;
+}
+
+int
+afr_get_split_brain_status (void *opaque)
 {
         gf_boolean_t      d_spb             = _gf_false;
         gf_boolean_t      m_spb             = _gf_false;
@@ -4727,7 +4733,15 @@ afr_get_split_brain_status (call_frame_t *frame, xlator_t *this, loc_t *loc)
         inode_t          *inode             = NULL;
         afr_private_t    *priv              = NULL;
         xlator_t         **children         = NULL;
-
+        call_frame_t     *frame             = NULL;
+        xlator_t         *this              = NULL;
+        loc_t            *loc               = NULL;
+        afr_spb_status_t *data              = NULL;
+
+        data     = opaque;
+        frame    = data->frame;
+        this     = frame->this;
+        loc      = data->loc;
         priv     = this->private;
         children = priv->children;
 
@@ -4777,14 +4791,20 @@ afr_get_split_brain_status (call_frame_t *frame, xlator_t *this, loc_t *loc)
                         goto out;
                 }
                 ret = dict_set_dynstr (dict, GF_AFR_SBRAIN_STATUS, status);
-                if (ret)
+                if (ret) {
+                        op_errno = -ret;
+                        ret = -1;
                         goto out;
+                }
         } else {
                 ret = dict_set_str (dict, GF_AFR_SBRAIN_STATUS,
                                     "The file is not under data or"
                                     " metadata split-brain");
-                if (ret)
+                if (ret) {
+                        op_errno = -ret;
+                        ret = -1;
                         goto out;
+                }
         }
 
         ret = 0;
diff --git a/xlators/cluster/afr/src/afr-inode-read.c b/xlators/cluster/afr/src/afr-inode-read.c
index 1ebe5e5caf8..6afd51dfb4e 100644
--- a/xlators/cluster/afr/src/afr-inode-read.c
+++ b/xlators/cluster/afr/src/afr-inode-read.c
@@ -1429,6 +1429,59 @@ afr_marker_populate_args (call_frame_t *frame, int type, int *gauge,
         return priv->child_count;
 }
 
+static int
+afr_handle_heal_xattrs (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                        const char *heal_op)
+{
+        int                     ret     = -1;
+        afr_spb_status_t       *data    = NULL;
+
+        if (!strcmp (heal_op, GF_HEAL_INFO)) {
+                afr_get_heal_info (frame, this, loc);
+                ret = 0;
+                goto out;
+        }
+
+        if (!strcmp (heal_op, GF_AFR_HEAL_SBRAIN)) {
+                afr_heal_splitbrain_file (frame, this, loc);
+                ret = 0;
+                goto out;
+        }
+
+        if (!strcmp (heal_op, GF_AFR_SBRAIN_STATUS)) {
+                data = GF_CALLOC (1, sizeof (*data), gf_afr_mt_spb_status_t);
+                if (!data) {
+                        ret = 1;
+                        goto out;
+                }
+                data->frame = frame;
+                data->loc = loc;
+                ret = synctask_new (this->ctx->env,
+                                    afr_get_split_brain_status,
+                                    afr_get_split_brain_status_cbk,
+                                    NULL, data);
+                if (ret) {
+                        gf_msg (this->name, GF_LOG_ERROR, 0,
+                                AFR_MSG_SPLIT_BRAIN_STATUS,
+                                "Failed to create"
+                                " synctask. Unable to fetch split-brain status"
+                                " for %s.", loc->name);
+                        ret = 1;
+                        goto out;
+                }
+                goto out;
+        }
+
+out:
+        if (ret == 1) {
+                AFR_STACK_UNWIND (getxattr, frame, -1, ENOMEM, NULL, NULL);
+                if (data)
+                        GF_FREE (data);
+                ret = 0;
+        }
+        return ret;
+}
+
 int32_t
 afr_getxattr (call_frame_t *frame, xlator_t *this,
               loc_t *loc, const char *name, dict_t *xdata)
@@ -1478,20 +1531,10 @@ afr_getxattr (call_frame_t *frame, xlator_t *this,
                                             afr_marker_populate_args) == 0)
                 return 0;
 
-        if (!strcmp (name, GF_HEAL_INFO)) {
-                afr_get_heal_info (frame, this, loc, xdata);
+        ret = afr_handle_heal_xattrs (frame, this, &local->loc, name);
+        if (ret == 0)
                 return 0;
-        }
 
-        if (!strcmp (name, GF_AFR_HEAL_SBRAIN)) {
-                afr_heal_splitbrain_file (frame, this, loc);
-                return 0;
-        }
-
-        if (!strcmp (name, GF_AFR_SBRAIN_STATUS)) {
-                afr_get_split_brain_status (frame, this, loc);
-                return 0;
-        }
         /*
          * Special xattrs which need responses from all subvols
          */
diff --git a/xlators/cluster/afr/src/afr-inode-write.c b/xlators/cluster/afr/src/afr-inode-write.c
index 84876503e22..18dd5529fb4 100644
--- a/xlators/cluster/afr/src/afr-inode-write.c
+++ b/xlators/cluster/afr/src/afr-inode-write.c
@@ -1342,6 +1342,7 @@ afr_handle_split_brain_commands (xlator_t *this, call_frame_t *frame,
                                 " synctask. Aborting split-brain choice set"
                                 " for %s", loc->name);
                         ret = 1;
+                        op_errno = ENOMEM;
                         goto out;
                 }
                 ret = 0;
@@ -1365,6 +1366,8 @@ out:
         /* key was correct but value was invalid when ret == 1 */
         if (ret == 1) {
                 AFR_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
+                if (data)
+                        GF_FREE (data);
                 ret = 0;
         }
         return ret;
diff --git a/xlators/cluster/afr/src/afr-mem-types.h b/xlators/cluster/afr/src/afr-mem-types.h
index a11063c1f25..fd484e4f1fd 100644
--- a/xlators/cluster/afr/src/afr-mem-types.h
+++ b/xlators/cluster/afr/src/afr-mem-types.h
@@ -44,6 +44,7 @@ enum gf_afr_mem_types_ {
 	gf_afr_mt_reply_t,
 	gf_afr_mt_subvol_healer_t,
 	gf_afr_mt_spbc_timeout_t,
+        gf_afr_mt_spb_status_t,
         gf_afr_mt_end
 };
 #endif
diff --git a/xlators/cluster/afr/src/afr-messages.h b/xlators/cluster/afr/src/afr-messages.h
index e4bf85b769f..4793413ed83 100644
--- a/xlators/cluster/afr/src/afr-messages.h
+++ b/xlators/cluster/afr/src/afr-messages.h
@@ -45,7 +45,7 @@
  */
 
 #define GLFS_COMP_BASE_AFR      GLFS_MSGID_COMP_AFR
-#define GLFS_NUM_MESSAGES       37
+#define GLFS_NUM_MESSAGES       38
 #define GLFS_MSGID_END          (GLFS_COMP_BASE_AFR + GLFS_NUM_MESSAGES + 1)
 
 #define glfs_msg_start_x GLFS_COMP_BASE_AFR, "Invalid: Start of messages"
@@ -333,6 +333,13 @@
 #define AFR_MSG_SELF_HEAL_FAILED                (GLFS_COMP_BASE_AFR + 37)
 
 
+/*!
+ * @messageid 108038
+ * @diagnosis
+ * @recommendedaction
+*/
+#define AFR_MSG_SPLIT_BRAIN_STATUS      (GLFS_COMP_BASE_AFR + 38)
+
 
 #define glfs_msg_end_x GLFS_MSGID_END, "Invalid: End of messages"
 #endif /* !_AFR_MESSAGES_H_ */
diff --git a/xlators/cluster/afr/src/afr.h b/xlators/cluster/afr/src/afr.h
index 008839a7312..870d26b7689 100644
--- a/xlators/cluster/afr/src/afr.h
+++ b/xlators/cluster/afr/src/afr.h
@@ -755,6 +755,11 @@ typedef struct afr_spbc_timeout {
         int          spb_child_index;
 } afr_spbc_timeout_t;
 
+typedef struct afr_spb_status {
+        call_frame_t *frame;
+        loc_t        *loc;
+} afr_spb_status_t;
+
 typedef struct afr_read_subvol_args {
         ia_type_t ia_type;
         uuid_t gfid;
@@ -1050,14 +1055,16 @@ gf_boolean_t
 afr_is_xattr_ignorable (char *key);
 
 int
-afr_get_heal_info (call_frame_t *frame, xlator_t *this, loc_t *loc,
-                   dict_t *xdata);
+afr_get_heal_info (call_frame_t *frame, xlator_t *this, loc_t *loc);
 
 int
 afr_heal_splitbrain_file(call_frame_t *frame, xlator_t *this, loc_t *loc);
 
 int
-afr_get_split_brain_status (call_frame_t *frame, xlator_t *this, loc_t *loc);
+afr_get_split_brain_status (void *opaque);
+
+int
+afr_get_split_brain_status_cbk (int ret, call_frame_t *frame, void *opaque);
 
 int
 afr_inode_split_brain_choice_set (inode_t *inode, xlator_t *this,