cluster/afr: Prioritize ENOSPC over other errors

Problem:
In a replicate/arbiter volume if file creations or writes fails on
quorum number of bricks and on one brick it is due to ENOSPC and
on other brick it fails for a different reason, it may fail with
errors other than ENOSPC in some cases.

Fix:
Prioritize ENOSPC over other lesser priority errors and do not set
op_errno in posix_gfid_set if op_ret is 0 to avoid receiving any
error_no which can be misinterpreted by __afr_dir_write_finalize().

Also removing the function afr_has_arbiter_fop_cbk_quorum() which
might consider a successful reply form a single brick as quorum
success in some cases, whereas we always need fop to be successful
on quorum number of bricks in arbiter configuration.

Change-Id: I106e267f8b9451f681022f1cccb410d9bc824c08
Fixes: #1254
Signed-off-by: karthik-us <ksubrahm@redhat.com>
(cherry picked from commit fa63b45ca5edf172b1b89b28b5db3c5129cc57b6)

4 files changed, 86 insertions, 48 deletions

diff --git a/tests/bugs/replicate/issue-1254-prioritize-enospc.t b/tests/bugs/replicate/issue-1254-prioritize-enospc.t
new file mode 100644
index 00000000000..fab94b71b27
--- /dev/null
+++ b/tests/bugs/replicate/issue-1254-prioritize-enospc.t
@@ -0,0 +1,80 @@
+#!/bin/bash
+
+. $(dirname $0)/../../include.rc
+. $(dirname $0)/../../volume.rc
+
+cleanup
+
+function create_bricks {
+    TEST truncate -s 100M $B0/brick0
+    TEST truncate -s 100M $B0/brick1
+    TEST truncate -s 20M $B0/brick2
+    LO1=`SETUP_LOOP $B0/brick0`
+    TEST [ $? -eq 0 ]
+    TEST MKFS_LOOP $LO1
+    LO2=`SETUP_LOOP $B0/brick1`
+    TEST [ $? -eq 0 ]
+    TEST MKFS_LOOP $LO2
+    LO3=`SETUP_LOOP $B0/brick2`
+    TEST [ $? -eq 0 ]
+    TEST MKFS_LOOP $LO3
+    TEST mkdir -p $B0/${V0}0 $B0/${V0}1 $B0/${V0}2
+    TEST MOUNT_LOOP $LO1 $B0/${V0}0
+    TEST MOUNT_LOOP $LO2 $B0/${V0}1
+    TEST MOUNT_LOOP $LO3 $B0/${V0}2
+}
+
+function create_files {
+        local i=1
+        while (true)
+        do
+                touch $M0/file$i
+                if [ -e $B0/${V0}2/file$i ];
+                then
+                        ((i++))
+                else
+                        break
+                fi
+        done
+}
+
+TESTS_EXPECTED_IN_LOOP=13
+
+#Arbiter volume: Check for ENOSPC when arbiter brick becomes full#
+TEST glusterd
+create_bricks
+TEST $CLI volume create $V0 replica 3 arbiter 1 $H0:$B0/${V0}{0,1,2}
+TEST $CLI volume start $V0
+TEST $CLI volume set $V0 performance.write-behind off
+TEST $GFS --volfile-server=$H0 --volfile-id=$V0 $M0
+
+create_files
+TEST kill_brick $V0 $H0 $B0/${V0}1
+error1=$(touch $M0/file-1 2>&1)
+EXPECT "No space left on device" echo $error1
+error2=$(mkdir $M0/dir-1 2>&1)
+EXPECT "No space left on device" echo $error2
+error3=$((echo "Test" > $M0/file-3) 2>&1)
+EXPECT "No space left on device" echo $error3
+
+cleanup
+
+#Replica-3 volume: Check for ENOSPC when one of the brick becomes full#
+#Keeping the third brick of lower size to simulate disk full scenario#
+TEST glusterd
+create_bricks
+TEST $CLI volume create $V0 replica 3 $H0:$B0/${V0}{0,1,2}
+TEST $CLI volume start $V0
+TEST $CLI volume set $V0 performance.write-behind off
+TEST $GFS --volfile-server=$H0 --volfile-id=$V0 $M0
+
+create_files
+TEST kill_brick $V0 $H0 $B0/${V0}1
+error1=$(touch $M0/file-1 2>&1)
+EXPECT "No space left on device" echo $error1
+error2=$(mkdir $M0/dir-1 2>&1)
+EXPECT "No space left on device" echo $error2
+error3=$((cat /dev/zero > $M0/file1) 2>&1)
+EXPECT "No space left on device" echo $error3
+
+cleanup
diff --git a/xlators/cluster/afr/src/afr-common.c b/xlators/cluster/afr/src/afr-common.c
index ab4acc13625..0fd080691f4 100644
--- a/xlators/cluster/afr/src/afr-common.c
+++ b/xlators/cluster/afr/src/afr-common.c
@@ -3051,7 +3051,7 @@ error:
  * others in that they must be given higher priority while
  * returning to the user.
  *
- * The hierarchy is ENODATA > ENOENT > ESTALE > others
+ * The hierarchy is ENODATA > ENOENT > ESTALE > ENOSPC others
  */
 
 int
@@ -3063,6 +3063,8 @@ afr_higher_errno(int32_t old_errno, int32_t new_errno)
         return ENOENT;
     if (old_errno == ESTALE || new_errno == ESTALE)
         return ESTALE;
+    if (old_errno == ENOSPC || new_errno == ENOSPC)
+        return ENOSPC;
 
     return new_errno;
 }
diff --git a/xlators/cluster/afr/src/afr-transaction.c b/xlators/cluster/afr/src/afr-transaction.c
index 78438f91331..c7a7d28fa7c 100644
--- a/xlators/cluster/afr/src/afr-transaction.c
+++ b/xlators/cluster/afr/src/afr-transaction.c
@@ -521,42 +521,6 @@ afr_compute_pre_op_sources(call_frame_t *frame, xlator_t *this)
                 local->transaction.pre_op_sources[j] = 0;
 }
 
-gf_boolean_t
-afr_has_arbiter_fop_cbk_quorum(call_frame_t *frame)
-{
-    afr_local_t *local = NULL;
-    afr_private_t *priv = NULL;
-    xlator_t *this = NULL;
-    gf_boolean_t fop_failed = _gf_false;
-    unsigned char *pre_op_sources = NULL;
-    int i = 0;
-
-    local = frame->local;
-    this = frame->this;
-    priv = this->private;
-    pre_op_sources = local->transaction.pre_op_sources;
-
-    /* If the fop failed on the brick, it is not a source. */
-    for (i = 0; i < priv->child_count; i++)
-        if (local->transaction.failed_subvols[i])
-            pre_op_sources[i] = 0;
-
-    switch (AFR_COUNT(pre_op_sources, priv->child_count)) {
-        case 1:
-            if (pre_op_sources[ARBITER_BRICK_INDEX])
-                fop_failed = _gf_true;
-            break;
-        case 0:
-            fop_failed = _gf_true;
-            break;
-    }
-
-    if (fop_failed)
-        return _gf_false;
-
-    return _gf_true;
-}
-
 void
 afr_txn_arbitrate_fop(call_frame_t *frame, xlator_t *this)
 {
@@ -971,12 +935,8 @@ afr_need_dirty_marking(call_frame_t *frame, xlator_t *this)
         priv->child_count)
         return _gf_false;
 
-    if (priv->arbiter_count) {
-        if (!afr_has_arbiter_fop_cbk_quorum(frame))
-            need_dirty = _gf_true;
-    } else if (!afr_has_fop_cbk_quorum(frame)) {
+    if (!afr_has_fop_cbk_quorum(frame))
         need_dirty = _gf_true;
-    }
 
     return need_dirty;
 }
@@ -1026,12 +986,8 @@ afr_handle_quorum(call_frame_t *frame, xlator_t *this)
      * no split-brain with the fix. The problem is eliminated completely.
      */
 
-    if (priv->arbiter_count) {
-        if (afr_has_arbiter_fop_cbk_quorum(frame))
-            return;
-    } else if (afr_has_fop_cbk_quorum(frame)) {
+    if (afr_has_fop_cbk_quorum(frame))
         return;
-    }
 
     if (afr_need_dirty_marking(frame, this))
         goto set_response;
diff --git a/xlators/storage/posix/src/posix-helpers.c b/xlators/storage/posix/src/posix-helpers.c
index b25855bbfbd..71c1a11ae98 100644
--- a/xlators/storage/posix/src/posix-helpers.c
+++ b/xlators/storage/posix/src/posix-helpers.c
@@ -1070,7 +1070,7 @@ verify_handle:
         ret = posix_handle_soft(this, path, loc, uuid_curr, &stat);
 
 out:
-    if (!(*op_errno))
+    if (ret && !(*op_errno))
         *op_errno = errno;
     return ret;
 }