afr: Modified book-keeping structures for entrylks

* There are upto 3 entry lockees that may be needed to perform
  entrylk'ing in posix dir-write operations.

* For eg, rmdir ("/a/b") needs to acquire locks on two entities,
  - entrylk ("/a", "b")
  - entrylk ("/a/b", null)

* Changed existing entrylk/rename/selfheal (entrylk) transactions
  to use the new book-keeping structures

* Fixed few issues in afr_trace_entry_lk{in,out} functions. Tracing is now
  aware of the new entry lockee structure.

Implementation notes:
* Changed 'cookie' sent in stack_wind to encode lockee_entity_no
  and subvol_no.

  cookie is a non-negative integer such that 0 <= cookie < replica_count,
  When more than one lock is being acquired across the subvolumes,
  cookie % replica_count gives the subvol_no
  cookie / replica_count gives the lockee_entity_no.

Change-Id: Idbf41803387a7d59a0f7fcb1453d91cea74da153
BUG: 765564
Signed-off-by: Krishnan Parthasarathi <kp@gluster.com>
Reviewed-on: http://review.gluster.org/2828
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>

1 files changed, 69 insertions, 0 deletions

diff --git a/tests/bugs/bug-765564.t b/tests/bugs/bug-765564.t
new file mode 100644
index 00000000000..cf00f520f14
--- /dev/null
+++ b/tests/bugs/bug-765564.t
@@ -0,0 +1,69 @@
+#!/bin/bash
+
+. $(dirname $0)/../include.rc
+
+TEST glusterd
+TEST pidof glusterd
+
+## Start and create a volume
+mkdir -p ${B0}/${V0}-0
+mkdir -p ${B0}/${V0}-1
+TEST $CLI volume create $V0 replica 2 $H0:$B0/${V0}-{0,1}
+
+TEST $CLI volume set $V0 performance.io-cache off;
+TEST $CLI volume set $V0 performance.write-behind off;
+TEST $CLI volume set $V0 performance.stat-prefetch off
+
+TEST $CLI volume start $V0;
+
+## Mount native
+TEST glusterfs --volfile-server=$H0 --volfile-id=$V0 $M0
+
+#returns success if 'olddir' is absent
+#'olddir' must be absent in both replicas
+function rm_succeeded () {
+    local dir1=$1
+    [[ -d $H0:$B0/${V0}-0/$dir1 || -d $H0:$B0/${V0}-1/$dir1 ]] && return 0
+    return 1
+}
+
+# returns successes if 'newdir' is present
+#'newdir' must be present in both replicas
+function mv_succeeded () {
+    local dir1=$1
+    [[ -d $H0:$B0/${V0}-0/$dir1 && -d $H0:$B0/${V0}-1/$dir1 ]] && return 1
+    return 0
+}
+
+# returns zero on success
+# Only one of rm and mv can succeed. This is captured by the XOR below
+
+function chk_backend_consistency(){
+    local dir1=$1
+    local dir2=$2
+    local rm_status=rm_succeeded $dir1
+    local mv_status=mv_succeeded $dir2
+    [[ ( $rm_status && ! $mv_status ) || ( ! $rm_status && $mv_status ) ]] && return 0
+    return 1
+}
+
+#concurrent removal/rename of dirs
+function rm_mv_correctness () {
+    ret=0
+    for i in {1..100}; do
+        mkdir $M0/"dir"$i
+        rmdir $M0/"dir"$i &
+        mv $M0/"dir"$i $M0/"adir"$i &
+        wait
+        tmp_ret=$(chk_backend_consistency "dir"$i "adir"$i)
+        (( ret += tmp_ret ))
+        rm -rf $M0/"dir"$i
+        rm -rf $M0/"adir"$i
+    done
+    return $ret
+}
+
+TEST rm_mv_correctness;
+TEST umount $M0;
+cleanup;
+