0 files changed, 0 insertions, 0 deletions
diff --git a/libglusterfs/src/gf-dirent.c b/libglusterfs/src/gf-dirent.c
index bb028c967..0cda83a27 100644
--- a/libglusterfs/src/gf-dirent.c
+++ b/libglusterfs/src/gf-dirent.c
@@ -83,6 +83,8 @@ gf_link_inodes_from_dirent (xlator_t *this, inode_t *parent,
                 if (entry->inode) {
                         link_inode = inode_link (entry->inode, parent,
                                                  entry->d_name, &entry->d_stat);
+			if (!link_inode)
+				continue;
                         inode_lookup (link_inode);
                         inode_unref (link_inode);
                 }
diff --git a/libglusterfs/src/glusterfs.h b/libglusterfs/src/glusterfs.h
index 31c46b74e..5ce0d6e70 100644
--- a/libglusterfs/src/glusterfs.h
+++ b/libglusterfs/src/glusterfs.h
@@ -132,7 +132,7 @@
 
 /* Index xlator related */
 #define GF_XATTROP_INDEX_GFID "glusterfs.xattrop_index_gfid"
-#define GF_BASE_INDICES_HOLDER_GFID "glusterfs.base_indicies_holder_gfid"
+#define GF_XATTROP_INDEX_COUNT "glusterfs.xattrop_index_count"
 
 #define GF_GFIDLESS_LOOKUP "gfidless-lookup"
 /* replace-brick and pump related internal xattrs */
diff --git a/libglusterfs/src/xlator.c b/libglusterfs/src/xlator.c
index f3df8e2ae..1bded6d3d 100644
--- a/libglusterfs/src/xlator.c
+++ b/libglusterfs/src/xlator.c
@@ -680,7 +680,9 @@ loc_copy_overload_parent (loc_t *dst, loc_t *src, inode_t *parent)
                         dst->name = strrchr (dst->path, '/');
                 if (dst->name)
                         dst->name++;
-        }
+        } else if (src->name) {
+		dst->name = src->name;
+	}
 
         ret = 0;
 out:
@@ -718,7 +720,9 @@ loc_copy (loc_t *dst, loc_t *src)
                         dst->name = strrchr (dst->path, '/');
                 if (dst->name)
                         dst->name++;
-        }
+        } else if (src->name) {
+		dst->name = src->name;
+	}
 
         ret = 0;
 out:
diff --git a/tests/basic/pump.t b/tests/basic/pump.t
index 3faf06f05..23bdc187d 100644
--- a/tests/basic/pump.t
+++ b/tests/basic/pump.t
@@ -22,7 +22,7 @@ done
 cd
 TEST umount $M0
 TEST $CLI volume replace-brick $V0 $H0:$B0/${V0}0 $H0:$B0/${V0}1 start
-EXPECT_WITHIN 60 "Y" gd_is_replace_brick_completed $H0 $V0 $H0:$B0/${V0}0 $H0:$B0/${V0}1
+EXPECT_WITHIN 600 "Y" gd_is_replace_brick_completed $H0 $V0 $H0:$B0/${V0}0 $H0:$B0/${V0}1
 TEST $CLI volume replace-brick $V0 $H0:$B0/${V0}0 $H0:$B0/${V0}1 commit
 TEST $CLI volume stop $V0
 TEST diff -r --exclude=.glusterfs $B0/${V0}0 $B0/${V0}1
diff --git a/tests/bugs/859927/repl.t b/tests/bugs/859927/repl.t
index 73c86e7be..856b057fb 100755
--- a/tests/bugs/859927/repl.t
+++ b/tests/bugs/859927/repl.t
@@ -33,20 +33,20 @@ TEST $CLI volume set $V0 cluster.data-self-heal-algorithm full
 EXPECT full volume_option $V0 cluster.data-self-heal-algorithm
 create_setup_for_self_heal $M0/a
 EXPECT_WITHIN 20 "1" afr_child_up_status $V0 0
-ls -l $file 2>&1 > /dev/null
+cat $file 2>&1 > /dev/null
 TEST cmp $B0/${V0}1/a $B0/${V0}2/a
 
 TEST $CLI volume set $V0 cluster.data-self-heal-algorithm diff
 EXPECT diff volume_option $V0 cluster.data-self-heal-algorithm
 create_setup_for_self_heal $M0/a
 EXPECT_WITHIN 20 "1" afr_child_up_status $V0 0
-ls -l $file 2>&1 > /dev/null
+cat $file 2>&1 > /dev/null
 TEST cmp $B0/${V0}1/a $B0/${V0}2/a
 
 TEST $CLI volume reset $V0 cluster.data-self-heal-algorithm
 create_setup_for_self_heal $M0/a
 EXPECT_WITHIN 20 "1" afr_child_up_status $V0 0
-ls -l $file 2>&1 > /dev/null
+cat $file 2>&1 > /dev/null
 TEST cmp $B0/${V0}1/a $B0/${V0}2/a
 
 TEST ! $CLI volume set $V0 cluster.data-self-heal-algorithm ""
diff --git a/tests/bugs/bug-1015990-rep.t b/tests/bugs/bug-1015990-rep.t
index f59bb2f75..bca0d7aff 100755
--- a/tests/bugs/bug-1015990-rep.t
+++ b/tests/bugs/bug-1015990-rep.t
@@ -35,7 +35,6 @@ for  i in  {1..100}; do echo "STRING" > $M0/File$i; done
 brick_2_sh_entries=$(count_sh_entries $B0/$V0"2")
 brick_4_sh_entries=$(count_sh_entries $B0/$V0"4")
 
-
 command_output=$(gluster volume heal $V0 statistics heal-count replica $H0:$B0/$V0"1")
 
 
diff --git a/tests/bugs/bug-1035576.t b/tests/bugs/bug-1035576.t
index 52d93dd87..938306a85 100644
--- a/tests/bugs/bug-1035576.t
+++ b/tests/bugs/bug-1035576.t
@@ -34,7 +34,8 @@ quota_limit_val1=$(get_hex_xattr trusted.glusterfs.quota.limit-set $B0/${V0}1/a)
 quota_size_val1=$(get_hex_xattr trusted.glusterfs.quota.size $B0/${V0}1/a)
 
 #Trigger entry,metadata self-heal
-TEST stat $M0/a
+TEST ls $M0/a
+
 quota_limit_val0=$(get_hex_xattr trusted.glusterfs.quota.limit-set $B0/${V0}0/a)
 quota_size_val0=$(get_hex_xattr trusted.glusterfs.quota.size $B0/${V0}0/a)
 
@@ -43,7 +44,7 @@ TEST [ $quota_limit_val0 == $quota_limit_val1 ]
 
 #Only entry, metadata self-heal is done quota size value should not be same
 TEST [ $quota_size_val0 != $quota_size_val1 ]
-TEST stat $M0/a/f
+TEST cat $M0/a/f
 
 #Now that data self-heal is done quota size value should be same
 quota_size_val0=$(get_hex_xattr trusted.glusterfs.quota.size $B0/${V0}0/a)
diff --git a/tests/bugs/bug-1037501.t b/tests/bugs/bug-1037501.t
index d11c788a0..596122a72 100755
--- a/tests/bugs/bug-1037501.t
+++ b/tests/bugs/bug-1037501.t
@@ -24,14 +24,6 @@ TEST $CLI volume create $V0 replica 3 $H0:$B0/${V0}-{0,1,2}
 EXPECT "$V0" volinfo_field $V0 'Volume Name';
 EXPECT 'Created' volinfo_field $V0 'Status';
 
-## Make sure io-cache and write-behind don't interfere.
-TEST $CLI volume set $V0 data-self-heal off;
-
-## Make sure automatic self-heal doesn't perturb our results.
-TEST $CLI volume set $V0 cluster.self-heal-daemon off
-
-TEST $CLI volume set $V0 background-self-heal-count 0
-
 ## Start volume and verify
 TEST $CLI volume start $V0;
 EXPECT 'Started' volinfo_field $V0 'Status';
@@ -48,206 +40,38 @@ TEST $CLI volume add-brick $V0 replica 4 $H0:$B0/$V0-3 force
 TEST $CLI volume add-brick $V0 replica 5 $H0:$B0/$V0-4 force
 TEST $CLI volume add-brick $V0 replica 6 $H0:$B0/$V0-5 force
 
-sleep 10
-
-TEST ls $M0/
-
-
-function compare()
-{
-	var=-1;
-	if [ $1 == $2 ]; then
-		var=0;
-	else
-		var=-1;
-	fi
-
-	echo $var
-}
-
-var2="000000000000000000000000"
-
-var1=`getfattr -d -m . $B0/$V0-0/File -e hex 2>&1 | grep "client-3"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-0/File -e hex 2>&1 | grep "client-4"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-0/File -e hex 2>&1 | grep "client-5"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-1/File -e hex 2>&1 | grep "client-3"`
-EXPECT "0" echo $?
-var3=`echo $var1| cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-1/File -e hex 2>&1 | grep "client-4"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-1/File -e hex 2>&1 | grep "client-5"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-2/File -e hex 2>&1 | grep "client-3"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-2/File -e hex 2>&1 | grep "client-4"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-2/File -e hex 2>&1 | grep "client-5"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-0/Dir -e hex 2>&1 | grep "client-3"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-0/Dir -e hex 2>&1 | grep "client-4"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-0/Dir -e hex 2>&1 | grep "client-5"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-1/Dir -e hex 2>&1 | grep "client-3"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-1/Dir -e hex 2>&1 | grep "client-4"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-1/Dir -e hex 2>&1 | grep "client-5"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-2/Dir -e hex 2>&1 | grep "client-3"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-2/Dir -e hex 2>&1 | grep "client-4"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-2/Dir -e hex 2>&1 | grep "client-5"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-
-var1=`getfattr -d -m . $B0/$V0-0/Link -e hex 2>&1 | grep "client-3"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-0/Link -e hex 2>&1 | grep "client-4"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-0/Link -e hex 2>&1 | grep "client-5"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-1/Link -e hex 2>&1 | grep "client-3"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-1/Link -e hex 2>&1 | grep "client-4"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-1/Link -e hex 2>&1 | grep "client-5"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-2/Link -e hex 2>&1 | grep "client-3"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-2/Link -e hex 2>&1 | grep "client-4"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-2/Link -e hex 2>&1 | grep "client-5"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-
-
-var1=`getfattr -d -m . $B0/$V0-0/FIFO -e hex 2>&1 | grep "client-3"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-0/FIFO -e hex 2>&1 | grep "client-4"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-0/FIFO -e hex 2>&1 | grep "client-5"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-1/FIFO -e hex 2>&1 | grep "client-3"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-1/FIFO -e hex 2>&1 | grep "client-4"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-1/FIFO -e hex 2>&1 | grep "client-5"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-2/FIFO -e hex 2>&1 | grep "client-3"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-2/FIFO -e hex 2>&1 | grep "client-4"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
-
-var1=`getfattr -d -m . $B0/$V0-2/FIFO -e hex 2>&1 | grep "client-5"`
-EXPECT "0" echo $?
-var3=`echo $var1 | cut -d x -f 2`
-EXPECT_NOT $var2 echo $var3
+sleep 5
+
+TEST gluster volume heal $V0 full
+
+sleep 5
+
+EXPECT 10 stat -c '%s' $B0/$V0-0/File
+EXPECT 10 stat -c '%s' $B0/$V0-1/File
+EXPECT 10 stat -c '%s' $B0/$V0-2/File
+EXPECT 10 stat -c '%s' $B0/$V0-3/File
+EXPECT 10 stat -c '%s' $B0/$V0-4/File
+EXPECT 10 stat -c '%s' $B0/$V0-5/File
+
+EXPECT 3 stat -c '%h' $B0/$V0-0/Link
+EXPECT 3 stat -c '%h' $B0/$V0-1/Link
+EXPECT 3 stat -c '%h' $B0/$V0-2/Link
+EXPECT 3 stat -c '%h' $B0/$V0-3/Link
+EXPECT 3 stat -c '%h' $B0/$V0-4/Link
+EXPECT 3 stat -c '%h' $B0/$V0-5/Link
+
+EXPECT 'directory' stat -c '%F' $B0/$V0-0/Dir
+EXPECT 'directory' stat -c '%F' $B0/$V0-1/Dir
+EXPECT 'directory' stat -c '%F' $B0/$V0-2/Dir
+EXPECT 'directory' stat -c '%F' $B0/$V0-3/Dir
+EXPECT 'directory' stat -c '%F' $B0/$V0-4/Dir
+EXPECT 'directory' stat -c '%F' $B0/$V0-5/Dir
+
+EXPECT 'fifo' stat -c '%F' $B0/$V0-0/FIFO
+EXPECT 'fifo' stat -c '%F' $B0/$V0-1/FIFO
+EXPECT 'fifo' stat -c '%F' $B0/$V0-2/FIFO
+EXPECT 'fifo' stat -c '%F' $B0/$V0-3/FIFO
+EXPECT 'fifo' stat -c '%F' $B0/$V0-4/FIFO
+EXPECT 'fifo' stat -c '%F' $B0/$V0-5/FIFO
 
 cleanup;
diff --git a/tests/bugs/bug-1058797.t b/tests/bugs/bug-1058797.t
index 2b80794cf..1e9f09af0 100644
--- a/tests/bugs/bug-1058797.t
+++ b/tests/bugs/bug-1058797.t
@@ -29,7 +29,7 @@ EXPECT "s" echo $setuid_bit1
 #Restart volume and do lookup from mount to trigger heal
 TEST $CLI volume start $V0 force
 EXPECT_WITHIN 20 "1" afr_child_up_status $V0 1
-TEST ls -l $M0/file
+TEST dd if=$M0/file of=/dev/null
 
 #Get file permissions from healed brick1 and verify that S_ISUID is indeed set
 file_permissions2=`ls -l $B0/brick1/file | awk '{print $1}' | cut -d. -f1 | cut -d- -f2,3,4,5,6`
diff --git a/tests/bugs/bug-767585-gfid.t b/tests/bugs/bug-767585-gfid.t
index 49cf7423f..41043a0b2 100755
--- a/tests/bugs/bug-767585-gfid.t
+++ b/tests/bugs/bug-767585-gfid.t
@@ -26,10 +26,9 @@ TEST setfattr -n trusted.gfid -v $gfid2 $B0/${V0}1/c
 
 sleep 2
 
-cd $M0
-TEST ls -l a
-TEST ls -l b
-TEST ls -l c
+TEST stat $M0/a
+TEST stat $M0/b
+TEST stat $M0/c
 
 TEST gf_get_gfid_xattr $B0/${V0}0/a
 TEST gf_get_gfid_xattr $B0/${V0}1/a
diff --git a/tests/bugs/bug-802417.t b/tests/bugs/bug-802417.t
index 314141f6b..b596df303 100755
--- a/tests/bugs/bug-802417.t
+++ b/tests/bugs/bug-802417.t
@@ -55,7 +55,7 @@ EXPECT_WITHIN 20 "1" afr_child_up_status $V0 0
 EXPECT_WITHIN 20 "1" afr_child_up_status $V0 1
 EXPECT_WITHIN 20 "1" afr_child_up_status $V0 2
 TEST kill_brick ${V0} ${H0} ${B0}/${V0}-2
-TEST ls -l ${M0}/a_file
+TEST dd if=${M0}/a_file of=/dev/null
 
 
 obs_path_0=${B0}/${V0}-0/a_file
@@ -67,31 +67,31 @@ tgt_xattr_1="trusted.afr.${V0}-client-1"
 tgt_xattr_2="trusted.afr.${V0}-client-2"
 
 actual=$(afr_get_changelog_xattr $obs_path_0 $tgt_xattr_0)
-EXPECT "0x000000000000000000000000" echo $actual
+EXPECT "0x000000000000000000000000|^\$" echo $actual
 
 actual=$(afr_get_changelog_xattr $obs_path_0 $tgt_xattr_1)
-EXPECT "0x000000000000000000000000" echo $actual
+EXPECT "0x000000000000000000000000|^\$" echo $actual
 
 actual=$(afr_get_changelog_xattr $obs_path_0 $tgt_xattr_2)
-EXPECT "0x000000020000000000000000" echo $actual
+EXPECT "0x000000030000000000000000" echo $actual
 
 actual=$(afr_get_changelog_xattr $obs_path_1 $tgt_xattr_0)
-EXPECT "0x000000000000000000000000" echo $actual
+EXPECT "0x000000000000000000000000|^\$" echo $actual
 
 actual=$(afr_get_changelog_xattr $obs_path_1 $tgt_xattr_1)
-EXPECT "0x000000000000000000000000" echo $actual
+EXPECT "0x000000000000000000000000|^\$" echo $actual
 
 actual=$(afr_get_changelog_xattr $obs_path_1 $tgt_xattr_2)
-EXPECT "0x000000020000000000000000" echo $actual
+EXPECT "0x000000010000000000000000" echo $actual
 
 actual=$(afr_get_changelog_xattr $obs_path_2 $tgt_xattr_0)
-EXPECT "0x000000000000000000000000" echo $actual
+EXPECT "0x000000000000000000000000|^\$" echo $actual
 
 actual=$(afr_get_changelog_xattr $obs_path_2 $tgt_xattr_1)
-EXPECT "0x000000000000000000000000" echo $actual
+EXPECT "0x000000000000000000000000|^\$" echo $actual
 
 actual=$(afr_get_changelog_xattr $obs_path_2 $tgt_xattr_2)
-EXPECT "0x000000000000000000000000" echo $actual
+EXPECT "0x000000000000000000000000|^\$" echo $actual
 
 if [ "$EXIT_EARLY" = "1" ]; then
 	exit 0;
diff --git a/tests/bugs/bug-830665.t b/tests/bugs/bug-830665.t
index 0073ff1d9..dd6f3ce2f 100755
--- a/tests/bugs/bug-830665.t
+++ b/tests/bugs/bug-830665.t
@@ -81,15 +81,17 @@ ls -l $N0 &> /dev/null;
 sleep 5;
 
 ## Force entry self-heal.
-find $N0 | xargs stat > /dev/null;
+TEST $CLI volume set $V0 cluster.self-heal-daemon on
+sleep 1
+TEST gluster volume heal $V0 full
 #ls -lR $N0 > /dev/null;
 
 ## Do NOT check through the NFS mount here. That will force a new self-heal
 ## check, but we want to test whether self-heal already happened.
 
 ## Make sure everything's in order on the recreated brick.
-EXPECT 'test_data' cat $B0/${V0}-0/a_file;
-EXPECT 'more_test_data' cat $B0/${V0}-0/a_dir/another_file;
+EXPECT_WITHIN 20 'test_data' cat $B0/${V0}-0/a_file;
+EXPECT_WITHIN 20 'more_test_data' cat $B0/${V0}-0/a_dir/another_file;
 
 if [ "$EXIT_EARLY" = "1" ]; then
 	exit 0;
diff --git a/tests/bugs/bug-853690.t b/tests/bugs/bug-853690.t
index 77a581f54..c2f82d103 100755
--- a/tests/bugs/bug-853690.t
+++ b/tests/bugs/bug-853690.t
@@ -66,7 +66,6 @@ TEST glusterfs --volfile=$B0/test.vol --attribute-timeout=0 --entry-timeout=0 $M
 # file sizes and immediate split-brain (EIO).
 TEST dd if=/dev/zero of=$M0/file bs=128k count=1
 TEST dd if=$M0/file of=/dev/null bs=128k count=1
-
 ########
 #
 # Test self-heal with short writes...
@@ -76,14 +75,11 @@ TEST dd if=$M0/file of=/dev/null bs=128k count=1
 # Cause a lookup and wait a few seconds for posterity. This self-heal also fails
 # due to a short write.
 TEST ls $M0/file
-
 # Verify the attributes on the healthy replica do not reflect consistency with
 # the other replica.
-TEST "getfattr -n trusted.afr.test-locks-0 $B0/test2/file --only-values > $B0/out1 2> /dev/null"
-TEST "getfattr -n trusted.afr.test-locks-1 $B0/test2/file --only-values > $B0/out2 2> /dev/null"
-TEST ! cmp $B0/out1 $B0/out2
+xa=`getfattr -n trusted.afr.test-locks-0 -e hex $B0/test2/file 2>&1 | grep = | cut -f2 -d=`
+EXPECT_NOT 0x000000000000000000000000 echo $xa
 
-TEST rm -f $B0/out1 $B0/out2
 TEST rm -f $M0/file
 TEST umount $M0
 
diff --git a/tests/bugs/bug-865825.t b/tests/bugs/bug-865825.t
index 6bb1c2348..8ee751864 100755
--- a/tests/bugs/bug-865825.t
+++ b/tests/bugs/bug-865825.t
@@ -2,6 +2,8 @@
 
 . $(dirname $0)/../include.rc
 
+cleanup;
+
 TEST glusterd
 TEST pidof glusterd
 TEST $CLI volume info;
@@ -28,6 +30,7 @@ EXPECT 'Created' volinfo_field $V0 'Status';
 ## Make sure io-cache and write-behind don't interfere.
 TEST $CLI volume set $V0 cluster.background-self-heal-count 0
 TEST $CLI volume set $V0 performance.io-cache off;
+TEST $CLI volume set $V0 performance.quick-read off;
 TEST $CLI volume set $V0 performance.write-behind off;
 TEST $CLI volume set $V0 performance.stat-prefetch off
 
@@ -54,19 +57,18 @@ setfattr -n trusted.afr.${V0}-client-2 -v $value $B0/${V0}-0/a_file
 setfattr -x trusted.afr.${V0}-client-2 $B0/${V0}-1/a_file
 echo "wrong_data" > $B0/${V0}-2/a_file
 
-## Remount and force a self-heal.
-TEST glusterfs --volfile-server=$H0 --volfile-id=$V0 $M0
-stat ${M0}/a_file > /dev/null
+gluster volume set $V0 cluster.self-heal-daemon on
+sleep 3
+gluster volume heal $V0 full
 
 ## Make sure brick 2 now has the correct contents.
-EXPECT "test_data" cat $B0/${V0}-2/a_file
+EXPECT_WITHIN 30 "test_data" cat $B0/${V0}-2/a_file
 
 if [ "$EXIT_EARLY" = "1" ]; then
 	exit 0;
 fi
 
 ## Finish up
-TEST umount $M0;
 TEST $CLI volume stop $V0;
 EXPECT 'Stopped' volinfo_field $V0 'Status';
 
diff --git a/tests/bugs/bug-873962.t b/tests/bugs/bug-873962.t
index b245cc3da..0281417f0 100755
--- a/tests/bugs/bug-873962.t
+++ b/tests/bugs/bug-873962.t
@@ -61,11 +61,12 @@ EXPECT_WITHIN 20 "1" afr_child_up_status $V0 0
 EXPECT_WITHIN 20 "1" afr_child_up_status $V0 1
 
 TEST glusterfs --entry-timeout=0 --attribute-timeout=0 -s $H0 --volfile-id=$V0 $M1 --direct-io-mode=enable
+
 #Files are in split-brain, so open should fail
 TEST ! cat $M0/a;
 TEST ! cat $M1/a;
-TEST ! cat $M0/b;
-TEST ! cat $M1/b;
+TEST cat $M0/b;
+TEST cat $M1/b;
 
 #Reset split-brain status
 TEST setfattr -n trusted.afr.$V0-client-1 -v 0x000000000000000000000000 $B0/${V0}1/a;
@@ -75,6 +76,7 @@ TEST setfattr -n trusted.afr.$V0-client-1 -v 0x000000000000000000000000 $B0/${V0
 EXPECT "2" cat $M0/a;
 # FAIL HERE - see comment about cluster.self-heal-background-count above.
 EXPECT "2" cat $M1/a;
+TEST dd if=$M0/b of=/dev/null bs=1M
 EXPECT "def" getfattr -n trusted.mdata --only-values $M0/b 2>/dev/null
 EXPECT "def" getfattr -n trusted.mdata --only-values $M1/b 2>/dev/null
 
@@ -90,8 +92,8 @@ TEST glusterfs --entry-timeout=0 --attribute-timeout=0 -s $H0 --volfile-id=$V0 $
 #Files are in split-brain, so open should fail
 TEST ! cat $M0/c
 TEST ! cat $M1/c
-TEST ! cat $M0/d
-TEST ! cat $M1/d
+TEST cat $M0/d
+TEST cat $M1/d
 
 TEST setfattr -n trusted.afr.$V0-client-1 -v 0x000000000000000000000000 $B0/${V0}1/c
 TEST setfattr -n trusted.afr.$V0-client-1 -v 0x000000000000000000000000 $B0/${V0}1/d
@@ -102,7 +104,4 @@ EXPECT "2" cat $M1/c
 EXPECT "1" cat $M0/d
 EXPECT "1" cat $M1/d
 
-#Check that the self-heal is not triggered.
-EXPECT "1" cat $B0/${V0}1/c
-EXPECT "abc" getfattr -n trusted.mdata --only-values $B0/${V0}1/d 2>/dev/null
 cleanup;
diff --git a/tests/bugs/bug-888174.t b/tests/bugs/bug-888174.t
index 4ea34645b..ef653f76d 100644
--- a/tests/bugs/bug-888174.t
+++ b/tests/bugs/bug-888174.t
@@ -38,10 +38,9 @@ TEST [ -z $inodelk_max_latency ]
 
 TEST dd of=$M0/a if=/dev/urandom bs=1M count=10 conv=fsync
 #Check for no trace of pending changelog. Flush should make sure of it.
-EXPECT "0x000000000000000000000000" afr_get_changelog_xattr $B0/r2_0/a trusted.afr.$V0-client-0
-EXPECT "0x000000000000000000000000" afr_get_changelog_xattr $B0/r2_0/a trusted.afr.$V0-client-1
-EXPECT "0x000000000000000000000000" afr_get_changelog_xattr $B0/r2_1/a trusted.afr.$V0-client-0
-EXPECT "0x000000000000000000000000" afr_get_changelog_xattr $B0/r2_1/a trusted.afr.$V0-client-1
+EXPECT "0x000000000000000000000000" afr_get_changelog_xattr $B0/r2_0/a trusted.afr.dirty
+EXPECT "0x000000000000000000000000" afr_get_changelog_xattr $B0/r2_1/a trusted.afr.dirty
+
 
 dd of=$M0/a if=/dev/urandom bs=1M count=1024 2>/dev/null &
 p=$!
@@ -51,15 +50,13 @@ TEST $CLI volume set $V0 performance.io-cache off
 TEST $CLI volume set $V0 performance.stat-prefetch off
 TEST $CLI volume set $V0 performance.read-ahead off
 
-kill -SIGTERM $p
+kill -TERM $p
 #wait for dd to exit
 wait  > /dev/null 2>&1
 
 #Goal is to check if there is permanent FOOL changelog
 sleep 5
-EXPECT "0x000000000000000000000000" afr_get_changelog_xattr $B0/r2_0/a trusted.afr.$V0-client-0
-EXPECT "0x000000000000000000000000" afr_get_changelog_xattr $B0/r2_0/a trusted.afr.$V0-client-1
-EXPECT "0x000000000000000000000000" afr_get_changelog_xattr $B0/r2_1/a trusted.afr.$V0-client-0
-EXPECT "0x000000000000000000000000" afr_get_changelog_xattr $B0/r2_1/a trusted.afr.$V0-client-1
+EXPECT "0x000000000000000000000000" afr_get_changelog_xattr $B0/r2_0/a trusted.afr.dirty
+EXPECT "0x000000000000000000000000" afr_get_changelog_xattr $B0/r2_1/a trusted.afr.dirty
 
 cleanup;
diff --git a/tests/bugs/bug-906646.t b/tests/bugs/bug-906646.t
index 0e6a3bcb6..b2cbf6bc3 100644
--- a/tests/bugs/bug-906646.t
+++ b/tests/bugs/bug-906646.t
@@ -84,7 +84,7 @@ TEST $CLI volume start $V0 force
 
 EXPECT_WITHIN 20 "1" afr_child_up_status $V0 `expr $brick_id - 1`
 
-stat $pth
+cat $pth >/dev/null
 
 # check backends - xattr should not be present anywhere
 EXPECT 1 xattr_query_check ${backend_paths_array[0]} "trusted.name"
diff --git a/tests/bugs/bug-913051.t b/tests/bugs/bug-913051.t
index 69e90cf66..9a59424f4 100644
--- a/tests/bugs/bug-913051.t
+++ b/tests/bugs/bug-913051.t
@@ -48,8 +48,8 @@ EXPECT "N" gf_check_file_opened_in_brick $V0 $H0 $B0/${V0}0 $B0/${V0}0/dir/b
 
 #attempt self-heal so that the files are created on brick-0
 
-TEST ls -l $M0/dir/a
-TEST ls -l $M0/dir/b
+TEST dd if=$M0/dir/a of=/dev/null bs=1M
+TEST dd if=$M0/dir/b of=/dev/null bs=1M
 
 #trigger writev for attempting open-fd-fix in afr
 TEST fd_write $wfd "open sesame"
diff --git a/tests/bugs/bug-913544.t b/tests/bugs/bug-913544.t
index 790bc0898..db28ca814 100644
--- a/tests/bugs/bug-913544.t
+++ b/tests/bugs/bug-913544.t
@@ -17,7 +17,7 @@ TEST touch a
 #simulate no-changelog data split-brain
 echo "abc" > $B0/${V0}1/a
 echo "abcd" > $B0/${V0}0/a
-TEST ! truncate -s 0 a
+TEST truncate -s 0 a
 TEST ls
 cd
 
diff --git a/tests/bugs/bug-918437-sh-mtime.t b/tests/bugs/bug-918437-sh-mtime.t
index 080956f51..11155ad16 100644
--- a/tests/bugs/bug-918437-sh-mtime.t
+++ b/tests/bugs/bug-918437-sh-mtime.t
@@ -38,7 +38,12 @@ TEST $CLI volume start $V0 force
 EXPECT_WITHIN 20 "1" afr_child_up_status $V0 0
 EXPECT_WITHIN 20 "1" afr_child_up_status $V0 1
 
-find $M0 | xargs stat 1>/dev/null
+TEST $CLI volume set $V0 cluster.self-heal-daemon on
+sleep 1
+TEST gluster volume heal $V0 full
+
+size=`stat -c '%s' /etc/passwd`
+EXPECT_WITHIN 60 $size stat -c '%s' $B0/gfs0/brick01/a
 
 TEST modify_atstamp1=$(get_mtime $B0/gfs0/brick01/a)
 TEST modify_atstamp2=$(get_mtime $B0/gfs0/brick02/a)
diff --git a/tests/bugs/bug-977797.t b/tests/bugs/bug-977797.t
index 08cdbe8f1..f2252159a 100755
--- a/tests/bugs/bug-977797.t
+++ b/tests/bugs/bug-977797.t
@@ -54,7 +54,7 @@ TEST chmod 757 $M0/a/file
 TEST $CLI volume start $V0 force
 EXPECT_WITHIN 20 "1" afr_child_up_status $V0 1;
 
-TEST ls   -l $M0/a/file
+TEST dd if=$M0/a/file of=/dev/null bs=1M
 
 b1c0dir=$(afr_get_specific_changelog_xattr $B0/$V0"1"/a \
           trusted.afr.$V0-client-0 "entry")
@@ -75,34 +75,15 @@ b2c0f=$(afr_get_specific_changelog_xattr $B0/$V0"2"/a/file \
 b2c1f=$(afr_get_specific_changelog_xattr $B0/$V0"2"/a/file \
         trusted.afr.$V0-client-1 "data")
 
-EXPECT "00000000" echo $b1c0f
-EXPECT "00000000" echo $b1c1f
-EXPECT "00000000" echo $b2c0f
-EXPECT "00000000" echo $b2c1f
-
-EXPECT "00000000" echo $b1c0dir
-EXPECT "00000000" echo $b1c1dir
-EXPECT "00000000" echo $b2c0dir
-EXPECT "00000000" echo $b2c1dir
-
-contains() {
-    string="$1"
-    substring="$2"
-    var="-1"
-    if test "${string#*$substring}" != "$string"
-    then
-        var="0"    # $substring is in $string
-    else
-        var="1"    # $substring is not in $string
-    fi
-    echo $var
-}
-
-var1=$(cat $M0/a/file 2>&1)
-var2="Input/output error"
-
-
-EXPECT "0" contains "$var1" "$var2"
+EXPECT "00000000|^$" echo $b1c0f
+EXPECT "00000000|^$" echo $b1c1f
+EXPECT "00000000|^$" echo $b2c0f
+EXPECT "00000000|^$" echo $b2c1f
+
+EXPECT "00000000|^$" echo $b1c0dir
+EXPECT "00000000|^$" echo $b1c1dir
+EXPECT "00000000|^$" echo $b2c0dir
+EXPECT "00000000|^$" echo $b2c1dir
 
 ## Finish up
 TEST $CLI volume stop $V0;
diff --git a/tests/volume.rc b/tests/volume.rc
index 5e2f95e76..9a06687cd 100644
--- a/tests/volume.rc
+++ b/tests/volume.rc
@@ -169,7 +169,7 @@ function check_option_help_presence {
 function afr_get_changelog_xattr {
         local file=$1
         local xkey=$2
-        getfattr -n $xkey -e hex $file 2>/dev/null | grep "client-" | cut -f2 -d'='
+        getfattr -n $xkey -e hex $file 2>/dev/null | grep "$xkey" | cut -f2 -d'='
 }
 
 function afr_get_pending_heal_count {
diff --git a/xlators/cluster/afr/src/Makefile.am b/xlators/cluster/afr/src/Makefile.am
index 35d18a6c0..ea5a90abb 100644
--- a/xlators/cluster/afr/src/Makefile.am
+++ b/xlators/cluster/afr/src/Makefile.am
@@ -2,24 +2,26 @@ xlator_LTLIBRARIES = afr.la pump.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/cluster
 
 afr_common_source = afr-dir-read.c afr-dir-write.c afr-inode-read.c \
-	afr-inode-write.c afr-open.c afr-transaction.c afr-self-heal-data.c \
-	afr-self-heal-common.c afr-self-heal-metadata.c afr-self-heal-entry.c \
-	afr-self-heal-algorithm.c afr-lk-common.c afr-self-heald.c \
+	afr-inode-write.c afr-open.c afr-transaction.c afr-lk-common.c \
+	afr-read-txn.c \
 	$(top_builddir)/xlators/lib/src/libxlator.c
 
+AFR_SELFHEAL_SOURCES = afr-self-heal-common.c afr-self-heal-data.c \
+	afr-self-heal-entry.c afr-self-heal-metadata.c afr-self-heald.c \
+	afr-self-heal-name.c
+
 afr_la_LDFLAGS = -module -avoid-version
-afr_la_SOURCES = $(afr_common_source) afr.c
+afr_la_SOURCES = $(afr_common_source) $(AFR_SELFHEAL_SOURCES) afr.c
 afr_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
 pump_la_LDFLAGS = -module -avoid-version
-pump_la_SOURCES =  $(afr_common_source) pump.c
+pump_la_SOURCES =  $(afr_common_source) $(AFR_SELFHEAL_SOURCES) pump.c
 pump_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
 noinst_HEADERS = afr.h afr-transaction.h afr-inode-write.h afr-inode-read.h \
-	afr-dir-read.h afr-dir-write.h afr-self-heal.h afr-self-heal-common.h \
-	afr-self-heal-algorithm.h pump.h afr-mem-types.h afr-common.c \
-	afr-self-heald.h $(top_builddir)/xlators/lib/src/libxlator.h \
-	$(top_builddir)/glusterfsd/src/glusterfsd.h
+	afr-dir-read.h afr-dir-write.h afr-self-heal.h afr-mem-types.h \
+	afr-common.c afr-self-heald.h pump.h \
+	$(top_builddir)/xlators/lib/src/libxlator.h
 
 AM_CPPFLAGS = $(GF_CPPFLAGS) \
 	-I$(top_srcdir)/libglusterfs/src -I$(top_srcdir)/xlators/lib/src \
@@ -31,7 +33,6 @@ CLEANFILES =
 
 uninstall-local:
 	rm -f $(DESTDIR)$(xlatordir)/replicate.so
-	rm -f $(DESTDIR)$(xlatordir)/pump.so
 
 install-data-hook:
 	ln -sf afr.so $(DESTDIR)$(xlatordir)/replicate.so
diff --git a/xlators/cluster/afr/src/afr-common.c b/xlators/cluster/afr/src/afr-common.c
index 224d30546..2bab0f853 100644
--- a/xlators/cluster/afr/src/afr-common.c
+++ b/xlators/cluster/afr/src/afr-common.c
@@ -45,787 +45,797 @@
 #include "afr-dir-write.h"
 #include "afr-transaction.h"
 #include "afr-self-heal.h"
-#include "afr-self-heal-common.h"
 #include "afr-self-heald.h"
-#include "pump.h"
 
-#define AFR_ICTX_OPENDIR_DONE_MASK     0x0000000100000000ULL
-#define AFR_ICTX_READ_CHILD_MASK       0x00000000FFFFFFFFULL
-#define AFR_STATISTICS_HISTORY_SIZE    50
-int
-afr_lookup_done_success_action (call_frame_t *frame, xlator_t *this,
-                                gf_boolean_t fail_conflict);
-void
-afr_children_copy (int32_t *dst, int32_t *src, unsigned int child_count)
-{
-        int     i = 0;
 
-        for (i = 0; i < child_count; i++)
-                dst[i] = src[i];
-}
-
-void
-afr_xattr_req_prepare (xlator_t *this, dict_t *xattr_req, const char *path)
+call_frame_t *
+afr_copy_frame (call_frame_t *base)
 {
-        int             i           = 0;
-        afr_private_t   *priv       = NULL;
-        int             ret         = 0;
+	afr_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	int op_errno = 0;
 
-        priv   = this->private;
+	frame = copy_frame (base);
+	if (!frame)
+		return NULL;
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local) {
+		AFR_STACK_DESTROY (frame);
+		return NULL;
+	}
 
-        for (i = 0; i < priv->child_count; i++) {
-                ret = dict_set_uint64 (xattr_req, priv->pending_key[i],
-                                       3 * sizeof(int32_t));
-                if (ret < 0)
-                        gf_log (this->name, GF_LOG_WARNING,
-                                "%s: Unable to set dict value for %s",
-                                path, priv->pending_key[i]);
-                /* 3 = data+metadata+entry */
-        }
-        ret = dict_set_int32 (xattr_req, GF_GFIDLESS_LOOKUP, 1);
-        if (ret) {
-                gf_log (this->name, GF_LOG_DEBUG, "%s: failed to set gfidless "
-                        "lookup", path);
-        }
+	return frame;
 }
 
+/*
+ * INODE CTX 64-bit VALUE FORMAT FOR SMALL (<= 16) SUBVOL COUNTS:
+ *
+ * |<----------   64bit   ------------>|
+ *  63           32 31    16 15       0
+ * |   EVENT_GEN   |  DATA  | METADATA |
+ *
+ *
+ *  METADATA (bit-0 .. bit-15): bitmap representing subvolumes from which
+ *                              metadata can be attempted to be read.
+ *
+ *                              bit-0 => priv->subvolumes[0]
+ *                              bit-1 => priv->subvolumes[1]
+ *                              ... etc. till bit-15
+ *
+ *  DATA (bit-16 .. bit-31): bitmap representing subvolumes from which data
+ *                           can be attempted to be read.
+ *
+ *                           bit-16 => priv->subvolumes[0]
+ *                           bit-17 => priv->subvolumes[1]
+ *                           ... etc. till bit-31
+ *
+ *  EVENT_GEN (bit-32 .. bit-63): event generation (i.e priv->event_generation)
+ *                                when DATA and METADATA was last updated.
+ *
+ *                                If EVENT_GEN is < priv->event_generation,
+ *                                or is 0, it means afr_inode_refresh() needs
+ *                                to be called to recalculate the bitmaps.
+ */
+
 int
-afr_lookup_xattr_req_prepare (afr_local_t *local, xlator_t *this,
-                              dict_t *xattr_req, loc_t *loc, void **gfid_req)
+__afr_inode_read_subvol_get_small (inode_t *inode, xlator_t *this,
+				   unsigned char *data, unsigned char *metadata,
+				   int *event_p)
 {
-        int     ret = -ENOMEM;
+	afr_private_t *priv = NULL;
+	int ret = -1;
+	uint16_t datamap = 0;
+	uint16_t metadatamap = 0;
+	uint32_t event = 0;
+	uint64_t val = 0;
+	int i = 0;
 
-        GF_ASSERT (gfid_req);
+	priv = this->private;
 
-        *gfid_req = NULL;
-        local->xattr_req = dict_new ();
-        if (!local->xattr_req)
-                goto out;
-        if (xattr_req)
-                dict_copy (xattr_req, local->xattr_req);
+	ret = __inode_ctx_get (inode, this, &val);
+	if (ret < 0)
+		return ret;
 
-        afr_xattr_req_prepare (this, local->xattr_req, loc->path);
-        ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_INODELK_COUNT, 0);
-        if (ret < 0) {
-                gf_log (this->name, GF_LOG_WARNING,
-                        "%s: Unable to set dict value for %s",
-                        loc->path, GLUSTERFS_INODELK_COUNT);
-        }
-        ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_ENTRYLK_COUNT, 0);
-        if (ret < 0) {
-                gf_log (this->name, GF_LOG_WARNING,
-                        "%s: Unable to set dict value for %s",
-                        loc->path, GLUSTERFS_ENTRYLK_COUNT);
-        }
+	metadatamap = (val & 0x000000000000ffff);
+	datamap =     (val & 0x00000000ffff0000) >> 16;
+	event =       (val & 0xffffffff00000000) >> 32;
 
-        ret = dict_set_uint32 (local->xattr_req, GLUSTERFS_PARENT_ENTRYLK, 0);
-        if (ret < 0) {
-                gf_log (this->name, GF_LOG_WARNING,
-                        "%s: Unable to set dict value for %s",
-                        loc->path, GLUSTERFS_PARENT_ENTRYLK);
-        }
+	for (i = 0; i < priv->child_count; i++) {
+		if (metadata)
+			metadata[i] = (metadatamap >> i) & 1;
+		if (data)
+			data[i] = (datamap >> i) & 1;
+	}
 
-        ret = dict_get_ptr (local->xattr_req, "gfid-req", gfid_req);
-        if (ret) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "%s: failed to get the gfid from dict", loc->path);
-                *gfid_req = NULL;
-        } else {
-                if (loc->parent != NULL)
-                        dict_del (local->xattr_req, "gfid-req");
-        }
-        ret = 0;
-out:
-        return ret;
+	if (event_p)
+		*event_p = event;
+	return ret;
 }
 
-void
-afr_lookup_save_gfid (uuid_t dst, void* new, const loc_t *loc)
-{
-        inode_t  *inode = NULL;
-
-        inode = loc->inode;
-        if (inode && !uuid_is_null (inode->gfid))
-                uuid_copy (dst, inode->gfid);
-        else if (!uuid_is_null (loc->gfid))
-                uuid_copy (dst, loc->gfid);
-        else if (new && !uuid_is_null (new))
-                uuid_copy (dst, new);
-}
 
 int
-afr_errno_count (int32_t *children, int *child_errno,
-                 unsigned int child_count, int32_t op_errno)
-{
-        int i = 0;
-        int errno_count = 0;
-        int child = 0;
+__afr_inode_read_subvol_set_small (inode_t *inode, xlator_t *this,
+				   unsigned char *data, unsigned char *metadata,
+				   int event)
+{
+	afr_private_t *priv = NULL;
+	uint16_t datamap = 0;
+	uint16_t metadatamap = 0;
+	uint64_t val = 0;
+	int i = 0;
+
+	priv = this->private;
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (data[i])
+			datamap |= (1 << i);
+		if (metadata[i])
+			metadatamap |= (1 << i);
+	}
 
-        for (i = 0; i < child_count; i++) {
-                if (children) {
-                        child = children[i];
-                        if (child == -1)
-                                break;
-                } else {
-                        child = i;
-                }
-                if (child_errno[child] == op_errno)
-                        errno_count++;
-        }
-        return errno_count;
-}
+	val = ((uint64_t) metadatamap) |
+		(((uint64_t) datamap) << 16) |
+		(((uint64_t) event) << 32);
 
-int32_t
-afr_set_dict_gfid (dict_t *dict, uuid_t gfid)
-{
-        int ret       = 0;
-        uuid_t *pgfid = NULL;
+	return __inode_ctx_set (inode, this, &val);
+}
 
-        GF_ASSERT (gfid);
 
-        pgfid = GF_CALLOC (1, sizeof (uuid_t), gf_common_mt_char);
-        if (!pgfid) {
-                ret = -1;
-                goto out;
-        }
+int
+__afr_inode_read_subvol_reset_small (inode_t *inode, xlator_t *this)
+{
+	int ret = -1;
+	uint16_t datamap = 0;
+	uint16_t metadatamap = 0;
+	uint32_t event = 0;
+	uint64_t val = 0;
 
-        uuid_copy (*pgfid, gfid);
+	ret = __inode_ctx_get (inode, this, &val);
+	(void) ret;
 
-        ret = dict_set_dynptr (dict, "gfid-req", pgfid, sizeof (uuid_t));
-        if (ret)
-                gf_log (THIS->name, GF_LOG_ERROR, "gfid set failed");
+	metadatamap = (val & 0x000000000000ffff) >> 0;
+	datamap =     (val & 0x00000000ffff0000) >> 16;
+	event = 0;
 
-out:
-        if (ret && pgfid)
-                GF_FREE (pgfid);
+	val = ((uint64_t) metadatamap) |
+		(((uint64_t) datamap) << 16) |
+		(((uint64_t) event) << 32);
 
-        return ret;
+	return __inode_ctx_set (inode, this, &val);
 }
 
-void
-afr_inode_ctx_destroy (afr_inode_ctx_t *ctx)
-{
-        if (!ctx)
-                return;
-        GF_FREE (ctx->fresh_children);
-        GF_FREE (ctx);
-}
 
-afr_inode_ctx_t*
-__afr_inode_ctx_get (inode_t *inode, xlator_t *this)
+int
+__afr_inode_read_subvol_get (inode_t *inode, xlator_t *this,
+			     unsigned char *data, unsigned char *metadata,
+			     int *event_p)
 {
-        int             ret      = 0;
-        uint64_t        ctx_addr = 0;
-        afr_inode_ctx_t *ctx     = NULL;
-        afr_private_t   *priv    = NULL;
+	afr_private_t *priv = NULL;
+	int ret = -1;
 
-        priv = this->private;
-        ret = __inode_ctx_get (inode, this, &ctx_addr);
-        if (ret < 0)
-                ctx_addr = 0;
-        if (ctx_addr != 0) {
-                ctx = (afr_inode_ctx_t*) (long) ctx_addr;
-                goto out;
-        }
-        ctx = GF_CALLOC (1, sizeof (*ctx),
-                         gf_afr_mt_inode_ctx_t);
-        if (!ctx)
-                goto fail;
-        ctx->fresh_children = GF_CALLOC (priv->child_count,
-                                         sizeof (*ctx->fresh_children),
-                                         gf_afr_mt_int32_t);
-        if (!ctx->fresh_children)
-                goto fail;
-        ret = __inode_ctx_put (inode, this, (uint64_t)ctx);
-        if (ret) {
-                gf_log_callingfn (this->name, GF_LOG_ERROR, "failed to "
-                                  "set the inode ctx (%s)",
-                                  uuid_utoa (inode->gfid));
-                goto fail;
-        }
+	priv = this->private;
 
-out:
-        return ctx;
+	if (priv->child_count <= 16)
+		ret = __afr_inode_read_subvol_get_small (inode, this, data,
+							 metadata, event_p);
+	else
+		/* TBD: allocate structure with array and read from it */
+		ret = -1;
 
-fail:
-        afr_inode_ctx_destroy (ctx);
-        return NULL;
+	return ret;
 }
 
-afr_inode_ctx_t*
-afr_inode_ctx_get (inode_t *inode, xlator_t *this)
+
+int
+__afr_inode_read_subvol_set (inode_t *inode, xlator_t *this, unsigned char *data,
+			     unsigned char *metadata, int event)
 {
-        afr_inode_ctx_t *ctx = NULL;
+	afr_private_t *priv = NULL;
+	int ret = -1;
 
-        LOCK (&inode->lock);
-        {
-                ctx = __afr_inode_ctx_get (inode, this);
-        }
-        UNLOCK (&inode->lock);
-        return ctx;
+	priv = this->private;
+
+	if (priv->child_count <= 16)
+		ret = __afr_inode_read_subvol_set_small (inode, this, data,
+							 metadata, event);
+	else
+		ret = -1;
+
+	return ret;
 }
 
-void
-afr_inode_get_ctx_params (xlator_t *this, inode_t *inode,
-                          afr_inode_params_t *params)
+
+int
+__afr_inode_read_subvol_reset (inode_t *inode, xlator_t *this)
 {
-        GF_ASSERT (inode);
-        GF_ASSERT (params);
+	afr_private_t *priv = NULL;
+	int ret = -1;
 
-        afr_inode_ctx_t *ctx = NULL;
-        afr_private_t   *priv = NULL;
-        int             i = 0;
-        int32_t         read_child = -1;
-        int32_t         *fresh_children = NULL;
+	priv = this->private;
 
-        priv = this->private;
-        LOCK (&inode->lock);
-        {
-                ctx = __afr_inode_ctx_get (inode, this);
-                if (!ctx)
-                        goto unlock;
-                switch (params->op) {
-                case AFR_INODE_GET_READ_CTX:
-                        fresh_children = params->u.read_ctx.children;
-                        read_child = (int32_t)(ctx->masks &
-                                               AFR_ICTX_READ_CHILD_MASK);
-                        params->u.read_ctx.read_child = read_child;
-                        if (!fresh_children)
-                                goto unlock;
-                        for (i = 0; i < priv->child_count; i++)
-                                fresh_children[i] = ctx->fresh_children[i];
-                        break;
-                case AFR_INODE_GET_OPENDIR_DONE:
-                        params->u.value = _gf_false;
-                        if (ctx->masks & AFR_ICTX_OPENDIR_DONE_MASK)
-                                params->u.value = _gf_true;
-                        break;
-                default:
-                        GF_ASSERT (0);
-                        break;
-                }
-        }
-unlock:
-        UNLOCK (&inode->lock);
+	if (priv->child_count <= 16)
+		ret = __afr_inode_read_subvol_reset_small (inode, this);
+	else
+		ret = -1;
+
+	return ret;
 }
 
-gf_boolean_t
-afr_is_split_brain (xlator_t *this, inode_t *inode)
+
+int
+afr_inode_read_subvol_get (inode_t *inode, xlator_t *this, unsigned char *data,
+			   unsigned char *metadata, int *event_p)
 {
-        afr_inode_ctx_t *ctx = NULL;
-        gf_boolean_t    spb  = _gf_false;
+	int ret = -1;
 
-        ctx = afr_inode_ctx_get (inode, this);
-        if (!ctx)
-                goto out;
-        if ((ctx->mdata_spb == SPB) || (ctx->data_spb == SPB))
-                spb = _gf_true;
-out:
-        return spb;
+	LOCK(&inode->lock);
+	{
+		ret = __afr_inode_read_subvol_get (inode, this, data,
+						   metadata, event_p);
+	}
+	UNLOCK(&inode->lock);
+
+	return ret;
 }
 
-gf_boolean_t
-afr_is_opendir_done (xlator_t *this, inode_t *inode)
+
+int
+afr_inode_read_subvol_set (inode_t *inode, xlator_t *this, unsigned char *data,
+			   unsigned char *metadata, int event)
 {
-        afr_inode_params_t params = {0};
+	int ret = -1;
+
+	LOCK(&inode->lock);
+	{
+		ret = __afr_inode_read_subvol_set (inode, this, data, metadata,
+						   event);
+	}
+	UNLOCK(&inode->lock);
 
-        params.op = AFR_INODE_GET_OPENDIR_DONE;
-        afr_inode_get_ctx_params (this, inode, &params);
-        return params.u.value;
+	return ret;
 }
 
-int32_t
-afr_inode_get_read_ctx (xlator_t *this, inode_t *inode, int32_t *fresh_children)
+
+int
+afr_inode_read_subvol_reset (inode_t *inode, xlator_t *this)
 {
-        afr_inode_params_t      params = {0};
+	int ret = -1;
+
+	LOCK(&inode->lock);
+	{
+		ret = __afr_inode_read_subvol_reset (inode, this);
+	}
+	UNLOCK(&inode->lock);
 
-        params.op = AFR_INODE_GET_READ_CTX;
-        params.u.read_ctx.children = fresh_children;
-        afr_inode_get_ctx_params (this, inode, &params);
-        return params.u.read_ctx.read_child;
+	return ret;
 }
 
-void
-afr_inode_ctx_set_read_child (afr_inode_ctx_t *ctx, int32_t read_child)
-{
-        uint64_t        remaining_mask = 0;
-        uint64_t        mask         = 0;
 
-        remaining_mask = (~AFR_ICTX_READ_CHILD_MASK & ctx->masks);
-        mask = (AFR_ICTX_READ_CHILD_MASK & read_child);
-        ctx->masks = remaining_mask | mask;
-}
+int
+afr_accused_fill (xlator_t *this, dict_t *xdata, unsigned char *accused,
+		  afr_transaction_type type)
+{
+	afr_private_t *priv = NULL;
+	int i = 0;
+	int idx = afr_index_for_transaction_type (type);
+	void *pending_raw = NULL;
+	int pending[3];
+	int ret = 0;
+
+	priv = this->private;
+
+	for (i = 0; i < priv->child_count; i++) {
+		ret = dict_get_ptr (xdata, priv->pending_key[i],
+				    &pending_raw);
+		if (ret) /* no pending flags */
+			continue;
+		memcpy (pending, pending_raw, sizeof(pending));
+
+		if (ntoh32 (pending[idx]))
+			accused[i] = 1;
+	}
 
-void
-afr_inode_ctx_set_read_ctx (afr_inode_ctx_t *ctx, int32_t read_child,
-                            int32_t *fresh_children, int32_t child_count)
-{
-        int             i            = 0;
-
-        afr_inode_ctx_set_read_child (ctx, read_child);
-        for (i = 0; i < child_count; i++) {
-                if (fresh_children)
-                        ctx->fresh_children[i] = fresh_children[i];
-                else
-                        ctx->fresh_children[i] = -1;
-        }
+	return 0;
 }
 
-void
-afr_inode_ctx_rm_stale_children (afr_inode_ctx_t *ctx, int32_t *stale_children,
-                                 int32_t child_count)
+
+int
+afr_accuse_smallfiles (xlator_t *this, struct afr_reply *replies,
+		       unsigned char *data_accused)
 {
-        int             i            = 0;
-        int32_t         read_child   = -1;
+	int i = 0;
+	afr_private_t *priv = NULL;
+	uint64_t maxsize = 0;
 
-        GF_ASSERT (stale_children);
-        for (i = 0; i < child_count; i++) {
-                if (stale_children[i] == -1)
-                        break;
-                afr_children_rm_child (ctx->fresh_children,
-                                       stale_children[i], child_count);
-        }
-        read_child = (int32_t)(ctx->masks & AFR_ICTX_READ_CHILD_MASK);
-        if (!afr_is_child_present (ctx->fresh_children, child_count,
-                                   read_child))
-                afr_inode_ctx_set_read_child (ctx, ctx->fresh_children[0]);
-}
+	priv = this->private;
 
-void
-afr_inode_ctx_set_opendir_done (afr_inode_ctx_t *ctx)
-{
-        uint64_t        remaining_mask = 0;
-        uint64_t        mask = 0;
+	for (i = 0; i < priv->child_count; i++) {
+		if (data_accused[i])
+			continue;
+		if (replies[i].poststat.ia_size > maxsize)
+			maxsize = replies[i].poststat.ia_size;
+	}
 
-        remaining_mask = (~AFR_ICTX_OPENDIR_DONE_MASK & ctx->masks);
-        mask = (0xFFFFFFFFFFFFFFFFULL & AFR_ICTX_OPENDIR_DONE_MASK);
-        ctx->masks = remaining_mask | mask;
+	for (i = 0; i < priv->child_count; i++) {
+		if (data_accused[i])
+			continue;
+		if (replies[i].poststat.ia_size < maxsize)
+			data_accused[i] = 1;
+	}
+
+	return 0;
 }
 
-void
-afr_inode_set_ctx_params (xlator_t *this, inode_t *inode,
-                          afr_inode_params_t *params)
-{
-        GF_ASSERT (inode);
-        GF_ASSERT (params);
 
-        afr_inode_ctx_t *ctx            = NULL;
-        afr_private_t   *priv           = NULL;
-        int32_t         read_child      = -1;
-        int32_t         *fresh_children = NULL;
-        int32_t         *stale_children = NULL;
+int
+afr_replies_interpret (call_frame_t *frame, xlator_t *this, inode_t *inode)
+{
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	struct afr_reply *replies = NULL;
+	int event_generation = 0;
+	int i = 0;
+	unsigned char *data_accused = NULL;
+	unsigned char *metadata_accused = NULL;
+	unsigned char *data_readable = NULL;
+	unsigned char *metadata_readable = NULL;
+	int ret = 0;
 
-        priv = this->private;
-        LOCK (&inode->lock);
-        {
-                ctx = __afr_inode_ctx_get (inode, this);
-                if (!ctx)
-                        goto unlock;
-                switch (params->op) {
-                case AFR_INODE_SET_READ_CTX:
-                        read_child = params->u.read_ctx.read_child;
-                        fresh_children = params->u.read_ctx.children;
-                        afr_inode_ctx_set_read_ctx (ctx, read_child,
-                                                    fresh_children,
-                                                    priv->child_count);
-                        break;
-                case AFR_INODE_RM_STALE_CHILDREN:
-                        stale_children = params->u.read_ctx.children;
-                        afr_inode_ctx_rm_stale_children (ctx,
-                                                         stale_children,
-                                                         priv->child_count);
-                        break;
-                case AFR_INODE_SET_OPENDIR_DONE:
-                        afr_inode_ctx_set_opendir_done (ctx);
-                        break;
-                default:
-                        GF_ASSERT (0);
-                        break;
-                }
-        }
-unlock:
-        UNLOCK (&inode->lock);
-}
+	local = frame->local;
+	priv = this->private;
+	replies = local->replies;
+	event_generation = local->event_generation;
+
+	data_accused = alloca0 (priv->child_count);
+	data_readable = alloca0 (priv->child_count);
+	metadata_accused = alloca0 (priv->child_count);
+	metadata_readable = alloca0 (priv->child_count);
+
+	for (i = 0; i < priv->child_count; i++) {
+		data_readable[i] = 1;
+		metadata_readable[i] = 1;
+	}
 
-void
-afr_set_split_brain (xlator_t *this, inode_t *inode, afr_spb_state_t mdata_spb,
-                     afr_spb_state_t data_spb)
-{
-        afr_inode_ctx_t *ctx = NULL;
+	for (i = 0; i < priv->child_count; i++) {
+		if (!replies[i].valid) {
+			data_readable[i] = 0;
+			metadata_readable[i] = 0;
+			continue;
+		}
+
+		if (replies[i].op_ret == -1) {
+			data_readable[i] = 0;
+			metadata_readable[i] = 0;
+			continue;
+		}
+
+		afr_accused_fill (this, replies[i].xdata, data_accused,
+				  (inode->ia_type == IA_IFDIR) ?
+				   AFR_ENTRY_TRANSACTION : AFR_DATA_TRANSACTION);
+
+		afr_accused_fill (this, replies[i].xdata,
+				  metadata_accused, AFR_METADATA_TRANSACTION);
+
+	}
 
-        ctx = afr_inode_ctx_get (inode, this);
-        if (mdata_spb != DONT_KNOW)
-                ctx->mdata_spb = mdata_spb;
-        if (data_spb != DONT_KNOW)
-                ctx->data_spb = data_spb;
+	if (inode->ia_type != IA_IFDIR)
+		afr_accuse_smallfiles (this, replies, data_accused);
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (data_accused[i]) {
+			data_readable[i] = 0;
+			ret = 1;
+		}
+		if (metadata_accused[i]) {
+			metadata_readable[i] = 0;
+			ret = 1;
+		}
+	}
+
+	afr_inode_read_subvol_set (inode, this, data_readable,
+				   metadata_readable, event_generation);
+	return ret;
 }
 
-void
-afr_set_opendir_done (xlator_t *this, inode_t *inode)
-{
-        afr_inode_params_t params = {0};
 
-        params.op = AFR_INODE_SET_OPENDIR_DONE;
-        afr_inode_set_ctx_params (this, inode, &params);
+
+int
+afr_refresh_selfheal_done (int ret, call_frame_t *heal, void *opaque)
+{
+	if (heal)
+		STACK_DESTROY (heal->root);
+	return 0;
 }
 
-void
-afr_inode_set_read_ctx (xlator_t *this, inode_t *inode, int32_t read_child,
-                        int32_t *fresh_children)
+int
+afr_inode_refresh_err (call_frame_t *frame, xlator_t *this)
 {
-        afr_inode_params_t params = {0};
-        afr_private_t      *priv  = NULL;
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	int i = 0;
+	int err = 0;
 
-        priv = this->private;
-        GF_ASSERT (read_child >= 0);
-        GF_ASSERT (fresh_children);
-        GF_ASSERT (afr_is_child_present (fresh_children, priv->child_count,
-                                         read_child));
-
-        params.op = AFR_INODE_SET_READ_CTX;
-        params.u.read_ctx.read_child     = read_child;
-        params.u.read_ctx.children = fresh_children;
-        afr_inode_set_ctx_params (this, inode, &params);
+	local = frame->local;
+	priv = this->private;
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (local->replies[i].valid && !local->replies[i].op_ret) {
+			err = 0;
+			goto ret;
+		}
+	}
+
+	err = afr_final_errno (local, priv);
+ret:
+	return -err;
 }
 
-void
-afr_inode_rm_stale_children (xlator_t *this, inode_t *inode,
-                             int32_t *stale_children)
+
+int
+afr_refresh_selfheal_wrap (void *opaque)
 {
-        afr_inode_params_t params = {0};
+	call_frame_t *frame = opaque;
+	afr_local_t *local = NULL;
+	xlator_t *this = NULL;
+	int err = 0;
+
+	local = frame->local;
+	this = frame->this;
 
-        GF_ASSERT (stale_children);
+	afr_selfheal (frame->this, local->refreshinode->gfid);
 
-        params.op = AFR_INODE_RM_STALE_CHILDREN;
-        params.u.read_ctx.children = stale_children;
-        afr_inode_set_ctx_params (this, inode, &params);
+	afr_selfheal_unlocked_discover (frame, local->refreshinode,
+					local->refreshinode->gfid,
+					local->replies);
+
+	afr_replies_interpret (frame, this, local->refreshinode);
+
+	err = afr_inode_refresh_err (frame, this);
+
+	afr_replies_wipe (local, this->private);
+
+	local->refreshfn (frame, this, err);
+
+	return 0;
 }
 
+
 gf_boolean_t
-afr_is_source_child (int32_t *sources, int32_t child_count, int32_t child)
+afr_selfheal_enabled (xlator_t *this)
 {
-        gf_boolean_t             source_xattrs = _gf_false;
+	afr_private_t *priv = NULL;
+	gf_boolean_t data = _gf_false;
 
-        GF_ASSERT (child < child_count);
+	priv = this->private;
 
-        if ((child >= 0) && (child < child_count) &&
-             sources[child]) {
-                source_xattrs = _gf_true;
-        }
-        return source_xattrs;
+	gf_string2boolean (priv->data_self_heal, &data);
+
+	return data || priv->metadata_self_heal || priv->entry_self_heal;
 }
 
-gf_boolean_t
-afr_is_child_present (int32_t *success_children, int32_t child_count,
-                      int32_t child)
+
+
+int
+afr_inode_refresh_done (call_frame_t *frame, xlator_t *this)
 {
-        gf_boolean_t             success_child = _gf_false;
-        int                      i = 0;
+	call_frame_t *heal = NULL;
+	afr_local_t *local = NULL;
+	int ret = 0;
+	int err = 0;
 
-        GF_ASSERT (child < child_count);
+	local = frame->local;
 
-        for (i = 0; i < child_count; i++) {
-                if (success_children[i] == -1)
-                        break;
-                if (child == success_children[i]) {
-                        success_child = _gf_true;
-                        break;
-                }
-        }
-        return success_child;
+	ret = afr_replies_interpret (frame, this, local->refreshinode);
+
+	err = afr_inode_refresh_err (frame, this);
+
+	afr_replies_wipe (local, this->private);
+
+	if (ret && afr_selfheal_enabled (this)) {
+		heal = copy_frame (frame);
+		if (heal)
+			heal->root->pid = -1;
+		ret = synctask_new (this->ctx->env, afr_refresh_selfheal_wrap,
+				    afr_refresh_selfheal_done, heal, frame);
+		if (ret)
+			goto refresh_done;
+	} else {
+	refresh_done:
+		local->refreshfn (frame, this, err);
+	}
+
+	return 0;
 }
 
-gf_boolean_t
-afr_is_read_child (int32_t *success_children, int32_t *sources,
-                   int32_t child_count, int32_t child)
+
+int
+afr_inode_refresh_subvol_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+			      int op_ret, int op_errno, inode_t *inode,
+			      struct iatt *buf, dict_t *xdata, struct iatt *par)
 {
-        gf_boolean_t             success_child = _gf_false;
-        gf_boolean_t             source        = _gf_false;
+	afr_local_t *local = NULL;
+	int call_child = (long) cookie;
+	int call_count = 0;
 
-        if (child < 0) {
-                return _gf_false;
-        }
+	local = frame->local;
 
-        GF_ASSERT (success_children);
-        GF_ASSERT (child_count > 0);
+	local->replies[call_child].valid = 1;
+	local->replies[call_child].op_ret = op_ret;
+	local->replies[call_child].op_errno = op_errno;
+	if (op_ret != -1) {
+		local->replies[call_child].poststat = *buf;
+		local->replies[call_child].postparent = *par;
+		local->replies[call_child].xdata = dict_ref (xdata);
+	}
 
-        success_child = afr_is_child_present (success_children, child_count,
-                                              child);
-        if (!success_child)
-                goto out;
-        if (NULL == sources) {
-                source = _gf_true;
-                goto out;
-        }
-        source = afr_is_source_child (sources, child_count, child);
-out:
-        return (success_child && source);
+	call_count = afr_frame_return (frame);
+
+	if (call_count == 0)
+		afr_inode_refresh_done (frame, this);
+
+	return 0;
 }
 
-int32_t
-afr_hash_child (int32_t *success_children, int32_t child_count,
-                unsigned int hmode, uuid_t gfid)
+
+int
+afr_inode_refresh_subvol (call_frame_t *frame, xlator_t *this, int i,
+			  inode_t *inode, dict_t *xdata)
 {
-        uuid_t  gfid_copy = {0,};
-        pid_t pid;
+	loc_t loc = {0, };
+	afr_private_t *priv = NULL;
 
-        if (!hmode) {
-                return -1;
-        }
+	priv = this->private;
 
-        if (gfid) {
-               uuid_copy(gfid_copy,gfid);
-        }
-        if (hmode > 1) {
-                /*
-                 * Why getpid?  Because it's one of the cheapest calls
-                 * available - faster than gethostname etc. - and returns a
-                 * constant-length value that's sure to be shorter than a UUID.
-                 * It's still very unlikely to be the same across clients, so
-                 * it still provides good mixing.  We're not trying for
-                 * perfection here.  All we need is a low probability that
-                 * multiple clients won't converge on the same subvolume.
-                 */
-                pid = getpid();
-                memcpy (gfid_copy, &pid, sizeof(pid));
-        }
+	loc.inode = inode;
+	uuid_copy (loc.gfid, inode->gfid);
 
-        return SuperFastHash((char *)gfid_copy,
-                             sizeof(gfid_copy)) % child_count;
+	STACK_WIND_COOKIE (frame, afr_inode_refresh_subvol_cbk,
+			   (void *) (long) i, priv->children[i],
+			   priv->children[i]->fops->lookup, &loc, xdata);
+	return 0;
 }
 
-/* If sources is NULL the xattrs are assumed to be of source for all
- * success_children.
- */
+
 int
-afr_select_read_child_from_policy (int32_t *success_children,
-                                   int32_t child_count, int32_t prev_read_child,
-                                   int32_t config_read_child, int32_t *sources,
-                                   unsigned int hmode, uuid_t gfid)
+afr_inode_refresh_do (call_frame_t *frame, xlator_t *this)
 {
-        int32_t                  read_child   = -1;
-        int                      i            = 0;
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	int call_count = 0;
+	int i = 0;
+	dict_t *xdata = NULL;
 
-        GF_ASSERT (success_children);
+	priv = this->private;
+	local = frame->local;
 
-        read_child = config_read_child;
-        if (afr_is_read_child (success_children, sources, child_count,
-                               read_child))
-                goto out;
+	afr_replies_wipe (local, priv);
 
-        read_child = prev_read_child;
-        if (afr_is_read_child (success_children, sources, child_count,
-                               read_child))
-                goto out;
+	xdata = dict_new ();
+	if (!xdata) {
+		afr_inode_refresh_done (frame, this);
+		return 0;
+	}
 
-        read_child = afr_hash_child (success_children, child_count,
-                                     hmode, gfid);
-        if (afr_is_read_child (success_children, sources, child_count,
-                               read_child)) {
-                goto out;
-        }
+	if (afr_xattr_req_prepare (this, xdata) != 0) {
+		dict_unref (xdata);
+		afr_inode_refresh_done (frame, this);
+		return 0;
+	}
 
-        for (i = 0; i < child_count; i++) {
-                read_child = success_children[i];
-                if (read_child < 0)
-                        break;
-                if (afr_is_read_child (success_children, sources, child_count,
-                                       read_child))
-                        goto out;
-        }
-        read_child = -1;
+	local->call_count = AFR_COUNT (local->child_up, priv->child_count);
 
-out:
-        return read_child;
+	call_count = local->call_count;
+	for (i = 0; i < priv->child_count; i++) {
+		if (!local->child_up[i])
+			continue;
+
+		afr_inode_refresh_subvol (frame, this, i, local->refreshinode,
+					  xdata);
+
+		if (!--call_count)
+			break;
+	}
+
+	dict_unref (xdata);
+
+	return 0;
 }
 
-/* This function should be used when all the success_children are sources
- */
-void
-afr_set_read_ctx_from_policy (xlator_t *this, inode_t *inode,
-                              int32_t *fresh_children, int32_t prev_read_child,
-                              int32_t config_read_child, uuid_t gfid)
+
+int
+afr_inode_refresh (call_frame_t *frame, xlator_t *this, inode_t *inode,
+		   afr_inode_refresh_cbk_t refreshfn)
 {
-        int                      read_child = -1;
-        afr_private_t            *priv = NULL;
+	afr_local_t *local = NULL;
 
-        priv = this->private;
-        read_child = afr_select_read_child_from_policy (fresh_children,
-                                                        priv->child_count,
-                                                        prev_read_child,
-                                                        config_read_child,
-                                                        NULL,
-                                                        priv->hash_mode, gfid);
-        if (read_child >= 0)
-                afr_inode_set_read_ctx (this, inode, read_child,
-                                        fresh_children);
+	local = frame->local;
+
+	local->refreshfn = refreshfn;
+
+	if (local->refreshinode) {
+		inode_unref (local->refreshinode);
+		local->refreshinode = NULL;
+	}
+
+	local->refreshinode = inode_ref (inode);
+
+	afr_inode_refresh_do (frame, this);
+
+	return 0;
 }
 
-/* afr_next_call_child ()
- * This is a common function used by all the read-type fops
- * This function should not be called with the inode's read_children array.
- * The fop's handler should make a copy of the inode's read_children,
- * preferred read_child into the local vars, because while this function is
- * in execution there is a chance for inode's read_ctx to change.
- */
-int32_t
-afr_next_call_child (int32_t *fresh_children, unsigned char *child_up,
-                     size_t child_count, int32_t *last_index,
-                     int32_t read_child)
+
+int
+afr_xattr_req_prepare (xlator_t *this, dict_t *xattr_req)
 {
-        int             next_index      = 0;
-        int32_t         next_call_child = -1;
+        int             i           = 0;
+        afr_private_t   *priv       = NULL;
+        int             ret         = 0;
 
-        GF_ASSERT (last_index);
+        priv   = this->private;
 
-        next_index = *last_index;
-retry:
-        next_index++;
-        if ((next_index >= child_count) ||
-           (fresh_children[next_index] == -1))
-                goto out;
-        if ((fresh_children[next_index] == read_child) ||
-           (!child_up[fresh_children[next_index]]))
-                goto retry;
-        *last_index = next_index;
-        next_call_child = fresh_children[next_index];
-out:
-        return next_call_child;
+        for (i = 0; i < priv->child_count; i++) {
+                ret = dict_set_uint64 (xattr_req, priv->pending_key[i],
+                                       AFR_NUM_CHANGE_LOGS * sizeof(int));
+                if (ret < 0)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "Unable to set dict value for %s",
+                                priv->pending_key[i]);
+                /* 3 = data+metadata+entry */
+        }
+        ret = dict_set_uint64 (xattr_req, AFR_DIRTY,
+			       AFR_NUM_CHANGE_LOGS * sizeof(int));
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "failed to set dirty "
+                        "query flag");
+        }
+
+	return ret;
 }
 
- /* This function should not be called with the inode's read_children array.
- * The fop's handler should make a copy of the inode's read_children,
- * preferred read_child into the local vars, because while this function is
- * in execution there is a chance for inode's read_ctx to change.
- */
-int32_t
-afr_get_call_child (xlator_t *this, unsigned char *child_up, int32_t read_child,
-                    int32_t *fresh_children,
-                    int32_t *call_child, int32_t *last_index)
+int
+afr_lookup_xattr_req_prepare (afr_local_t *local, xlator_t *this,
+                              dict_t *xattr_req, loc_t *loc)
 {
-        int             ret   = 0;
-        afr_private_t   *priv = NULL;
-        int             i     = 0;
-
-        GF_ASSERT (child_up);
-        GF_ASSERT (call_child);
-        GF_ASSERT (last_index);
-        GF_ASSERT (fresh_children);
+        int     ret = -ENOMEM;
 
-        if (read_child < 0) {
-                ret = -EIO;
+        local->xattr_req = dict_new ();
+        if (!local->xattr_req)
                 goto out;
-        }
-        priv = this->private;
-        *call_child = -1;
-        *last_index = -1;
+        if (xattr_req)
+                dict_copy (xattr_req, local->xattr_req);
 
-        if (child_up[read_child]) {
-                *call_child = read_child;
-        } else {
-                for (i = 0; i < priv->child_count; i++) {
-                        if (fresh_children[i] == -1)
-                                break;
-                        if (child_up[fresh_children[i]]) {
-                                *call_child = fresh_children[i];
-                                ret = 0;
-                                break;
-                        }
-                }
+        ret = afr_xattr_req_prepare (this, local->xattr_req);
+	if (ret < 0) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"%s: Unable to prepare xattr_req", loc->path);
+	}
 
-                if (*call_child == -1) {
-                        ret = -ENOTCONN;
-                        goto out;
-                }
+        ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_INODELK_COUNT, 0);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: Unable to set dict value for %s",
+                        loc->path, GLUSTERFS_INODELK_COUNT);
+        }
+        ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_ENTRYLK_COUNT, 0);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: Unable to set dict value for %s",
+                        loc->path, GLUSTERFS_ENTRYLK_COUNT);
+        }
 
-                *last_index = i;
+        ret = dict_set_uint32 (local->xattr_req, GLUSTERFS_PARENT_ENTRYLK, 0);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: Unable to set dict value for %s",
+                        loc->path, GLUSTERFS_PARENT_ENTRYLK);
         }
+
+        ret = 0;
 out:
-        gf_log (this->name, GF_LOG_DEBUG, "Returning %d, call_child: %d, "
-                "last_index: %d", ret, *call_child, *last_index);
         return ret;
 }
 
-void
-afr_reset_xattr (dict_t **xattr, unsigned int child_count)
+
+int
+afr_hash_child (inode_t *inode, int32_t child_count, int hashmode)
 {
-        unsigned int i = 0;
+        uuid_t gfid_copy = {0,};
+        pid_t pid;
 
-        if (!xattr)
-                goto out;
-        for (i = 0; i < child_count; i++) {
-                if (xattr[i]) {
-                        dict_unref (xattr[i]);
-                        xattr[i] = NULL;
-                }
+        if (!hashmode) {
+                return -1;
         }
-out:
-        return;
-}
 
-void
-afr_xattr_array_destroy (dict_t **xattr, unsigned int child_count)
-{
-        afr_reset_xattr (xattr, child_count);
-        GF_FREE (xattr);
-}
+        if (inode) {
+               uuid_copy (gfid_copy, inode->gfid);
+        }
 
-void
-afr_local_sh_cleanup (afr_local_t *local, xlator_t *this)
-{
-        afr_self_heal_t *sh = NULL;
-        afr_private_t   *priv = NULL;
+        if (hashmode > 1) {
+                /*
+                 * Why getpid?  Because it's one of the cheapest calls
+                 * available - faster than gethostname etc. - and returns a
+                 * constant-length value that's sure to be shorter than a UUID.
+                 * It's still very unlikely to be the same across clients, so
+                 * it still provides good mixing.  We're not trying for
+                 * perfection here.  All we need is a low probability that
+                 * multiple clients won't converge on the same subvolume.
+                 */
+                pid = getpid();
+                memcpy (gfid_copy, &pid, sizeof(pid));
+        }
 
-        sh = &local->self_heal;
-        priv = this->private;
+        return SuperFastHash((char *)gfid_copy,
+                             sizeof(gfid_copy)) % child_count;
+}
 
-        if (sh->data_sh_info && strcmp (sh->data_sh_info, ""))
-                GF_FREE (sh->data_sh_info);
 
-        if (sh->metadata_sh_info && strcmp (sh->metadata_sh_info, ""))
-                GF_FREE (sh->metadata_sh_info);
+int
+afr_read_subvol_select_by_policy (inode_t *inode, xlator_t *this,
+				  unsigned char *readable)
+{
+	afr_private_t *priv = NULL;
+	int read_subvol = -1;
+	int i = 0;
 
-        GF_FREE (sh->buf);
+	priv = this->private;
 
-        GF_FREE (sh->parentbufs);
+	/* first preference - explicitly specified or local subvolume */
+	if (priv->read_child >= 0 && readable[priv->read_child])
+		return priv->read_child;
 
-        if (sh->inode)
-                inode_unref (sh->inode);
+	/* second preference - use hashed mode */
+	read_subvol = afr_hash_child (inode, priv->child_count,
+				      priv->hash_mode);
+	if (read_subvol >= 0 && readable[read_subvol])
+		return read_subvol;
 
-        afr_xattr_array_destroy (sh->xattr, priv->child_count);
+	for (i = 0; i < priv->child_count; i++) {
+		if (readable[i])
+			return i;
+	}
 
-        GF_FREE (sh->child_errno);
+	/* no readable subvolumes, either split brain or all subvols down */
 
-        afr_matrix_cleanup (sh->pending_matrix, priv->child_count);
-        afr_matrix_cleanup (sh->delta_matrix, priv->child_count);
+	return -1;
+}
 
-        GF_FREE (sh->sources);
 
-        GF_FREE (sh->success);
+int
+afr_inode_read_subvol_type_get (inode_t *inode, xlator_t *this,
+				unsigned char *readable, int *event_p,
+				int type)
+{
+	int ret = -1;
 
-        GF_FREE (sh->locked_nodes);
+	if (type == AFR_METADATA_TRANSACTION)
+		ret = afr_inode_read_subvol_get (inode, this, 0, readable,
+						 event_p);
+	else
+		ret = afr_inode_read_subvol_get (inode, this, readable, 0,
+						 event_p);
+	return ret;
+}
 
-        if (sh->healing_fd) {
-                fd_unref (sh->healing_fd);
-                sh->healing_fd = NULL;
-        }
 
-        GF_FREE ((char *)sh->linkname);
+int
+afr_read_subvol_get (inode_t *inode, xlator_t *this, int *subvol_p,
+		     int *event_p, afr_transaction_type type)
+{
+	afr_private_t *priv = NULL;
+	unsigned char *data_readable = NULL;
+	unsigned char *metadata_readable = NULL;
+	unsigned char *readable = NULL;
+	unsigned char *intersection = NULL;
+	int subvol = -1;
+	int event = 0;
 
-        GF_FREE (sh->success_children);
+	priv = this->private;
 
-        GF_FREE (sh->fresh_children);
+	readable = alloca0 (priv->child_count);
+	data_readable = alloca0 (priv->child_count);
+	metadata_readable = alloca0 (priv->child_count);
+	intersection = alloca0 (priv->child_count);
 
-        GF_FREE (sh->fresh_parent_dirs);
+	afr_inode_read_subvol_type_get (inode, this, readable, &event, type);
 
-        loc_wipe (&sh->parent_loc);
-        loc_wipe (&sh->lookup_loc);
+	afr_inode_read_subvol_get (inode, this, data_readable, metadata_readable,
+				   &event);
 
-        GF_FREE (sh->checksum);
+	AFR_INTERSECT (intersection, data_readable, metadata_readable,
+		       priv->child_count);
 
-        GF_FREE (sh->write_needed);
-        if (sh->healing_fd)
-                fd_unref (sh->healing_fd);
+	if (AFR_COUNT (intersection, priv->child_count) > 0)
+		subvol = afr_read_subvol_select_by_policy (inode, this,
+							   intersection);
+	else
+		subvol = afr_read_subvol_select_by_policy (inode, this,
+							   readable);
+	if (subvol_p)
+		*subvol_p = subvol;
+	if (event_p)
+		*event_p = event;
+	return subvol;
 }
 
 
@@ -838,8 +848,6 @@ afr_local_transaction_cleanup (afr_local_t *local, xlator_t *this)
         priv = this->private;
 
         afr_matrix_cleanup (local->pending, priv->child_count);
-        afr_matrix_cleanup (local->transaction.txn_changelog,
-                            priv->child_count);
 
         GF_FREE (local->internal_lock.locked_nodes);
 
@@ -860,7 +868,25 @@ afr_local_transaction_cleanup (afr_local_t *local, xlator_t *this)
         loc_wipe (&local->transaction.parent_loc);
         loc_wipe (&local->transaction.new_parent_loc);
 
-        GF_FREE (local->transaction.postop_piggybacked);
+}
+
+
+void
+afr_replies_wipe (afr_local_t *local, afr_private_t *priv)
+{
+	int i;
+
+	if (!local->replies)
+		return;
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (local->replies[i].xdata) {
+			dict_unref (local->replies[i].xdata);
+			local->replies[i].xdata = NULL;
+		}
+	}
+
+	memset (local->replies, 0, sizeof(*local->replies) * priv->child_count);
 }
 
 
@@ -872,7 +898,7 @@ afr_local_cleanup (afr_local_t *local, xlator_t *this)
         if (!local)
                 return;
 
-        afr_local_sh_cleanup (local, this);
+	syncbarrier_destroy (&local->barrier);
 
         afr_local_transaction_cleanup (local, this);
 
@@ -890,40 +916,26 @@ afr_local_cleanup (afr_local_t *local, xlator_t *this)
         if (local->dict)
                 dict_unref (local->dict);
 
+	afr_replies_wipe (local, priv);
 	GF_FREE(local->replies);
 
         GF_FREE (local->child_up);
 
-        GF_FREE (local->child_errno);
+        GF_FREE (local->read_attempted);
 
-        GF_FREE (local->fresh_childre