cluster/afr: Sparse file self-heal canges

- Fix boundary condition for offset
- Honour data-self-heal-algorithm option
- Added tests for sparse file self-healing

Change-Id: I14bb1c9d04118a3df4072f962fc8f2f197391d95
BUG: 1080707
Signed-off-by: Pranith Kumar K <pkarampu@redhat.com>
Reviewed-on: http://review.gluster.org/7339
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>

3 files changed, 157 insertions, 11 deletions

diff --git a/tests/basic/afr/sparse-file-self-heal.t b/tests/basic/afr/sparse-file-self-heal.t
new file mode 100644
index 00000000000..9b795c331cf
--- /dev/null
+++ b/tests/basic/afr/sparse-file-self-heal.t
@@ -0,0 +1,121 @@
+#!/bin/bash
+
+#This file checks if self-heal of files with holes is working properly or not
+#bigger is 2M, big is 1M, small is anything less
+. $(dirname $0)/../../include.rc
+. $(dirname $0)/../../volume.rc
+
+cleanup;
+
+TEST glusterd
+TEST pidof glusterd
+TEST $CLI volume create $V0 replica 2 $H0:$B0/${V0}{0,1}
+TEST $CLI volume set $V0 data-self-heal-algorithm full
+TEST $CLI volume start $V0
+
+TEST glusterfs --volfile-id=/$V0 --volfile-server=$H0 $M0 --attribute-timeout=0 --entry-timeout=0
+TEST dd if=/dev/urandom of=$M0/small count=1 bs=1M
+TEST dd if=/dev/urandom of=$M0/bigger2big count=1 bs=2M
+TEST dd if=/dev/urandom of=$M0/big2bigger count=1 bs=1M
+
+TEST kill_brick $V0 $H0 $B0/${V0}0
+
+#File with >128k size hole
+TEST truncate -s 1M $M0/big
+big_md5sum=$(md5sum $M0/big | awk '{print $1}')
+
+#File with <128k hole
+TEST truncate -s 0 $M0/small
+TEST truncate -s 64k $M0/small
+small_md5sum=$(md5sum $M0/small | awk '{print $1}')
+
+#Bigger file truncated to big size hole.
+TEST truncate -s 0 $M0/bigger2big
+TEST truncate -s 1M $M0/bigger2big
+bigger2big_md5sum=$(md5sum $M0/bigger2big | awk '{print $1}')
+
+#Big file truncated to Bigger size hole
+TEST truncate -s 2M $M0/big2bigger
+big2bigger_md5sum=$(md5sum $M0/big2bigger | awk '{print $1}')
+
+$CLI volume start $V0 force
+EXPECT_WITHIN 20 "1" afr_child_up_status $V0 0
+EXPECT_WITHIN 20 "Y" glustershd_up_status
+EXPECT_WITHIN 20 "1" afr_child_up_status_in_shd $V0 0
+EXPECT_WITHIN 20 "1" afr_child_up_status_in_shd $V0 1
+TEST gluster volume heal $V0 full
+EXPECT_WITHIN 20 "0" afr_get_pending_heal_count $V0
+
+big_md5sum_0=$(md5sum $B0/${V0}0/big | awk '{print $1}')
+small_md5sum_0=$(md5sum $B0/${V0}0/small | awk '{print $1}')
+bigger2big_md5sum_0=$(md5sum $B0/${V0}0/bigger2big | awk '{print $1}')
+big2bigger_md5sum_0=$(md5sum $B0/${V0}0/big2bigger | awk '{print $1}')
+
+EXPECT $big_md5sum echo $big_md5sum_0
+EXPECT $small_md5sum echo $small_md5sum_0
+EXPECT $big2bigger_md5sum echo $big2bigger_md5sum_0
+EXPECT $bigger2big_md5sum echo $bigger2big_md5sum_0
+
+
+EXPECT "1" has_holes $B0/${V0}0/big
+#Because self-heal writes the final chunk hole should not be there for
+#files < 128K
+EXPECT "0" has_holes $B0/${V0}0/small
+# Since source is smaller than sink, self-heal does blind copy so no holes will
+# be present
+EXPECT "0" has_holes $B0/${V0}0/bigger2big
+EXPECT "1" has_holes $B0/${V0}0/big2bigger
+
+TEST rm -f $M0/*
+
+#check the same tests with diff self-heal
+TEST $CLI volume set $V0 data-self-heal-algorithm diff
+
+TEST dd if=/dev/urandom of=$M0/small count=1 bs=1M
+TEST dd if=/dev/urandom of=$M0/big2bigger count=1 bs=1M
+TEST dd if=/dev/urandom of=$M0/bigger2big count=1 bs=2M
+
+TEST kill_brick $V0 $H0 $B0/${V0}0
+
+#File with >128k size hole
+TEST truncate -s 1M $M0/big
+big_md5sum=$(md5sum $M0/big | awk '{print $1}')
+
+#File with <128k hole
+TEST truncate -s 0 $M0/small
+TEST truncate -s 64k $M0/small
+small_md5sum=$(md5sum $M0/small | awk '{print $1}')
+
+#Bigger file truncated to big size hole
+TEST truncate -s 0 $M0/bigger2big
+TEST truncate -s 1M $M0/bigger2big
+bigger2big_md5sum=$(md5sum $M0/bigger2big | awk '{print $1}')
+
+#Big file truncated to Bigger size hole
+TEST truncate -s 2M $M0/big2bigger
+big2bigger_md5sum=$(md5sum $M0/big2bigger | awk '{print $1}')
+
+$CLI volume start $V0 force
+EXPECT_WITHIN 20 "1" afr_child_up_status $V0 0
+EXPECT_WITHIN 20 "Y" glustershd_up_status
+EXPECT_WITHIN 20 "1" afr_child_up_status_in_shd $V0 0
+EXPECT_WITHIN 20 "1" afr_child_up_status_in_shd $V0 1
+TEST gluster volume heal $V0 full
+EXPECT_WITHIN 20 "0" afr_get_pending_heal_count $V0
+
+big_md5sum_0=$(md5sum $B0/${V0}0/big | awk '{print $1}')
+small_md5sum_0=$(md5sum $B0/${V0}0/small | awk '{print $1}')
+bigger2big_md5sum_0=$(md5sum $B0/${V0}0/bigger2big | awk '{print $1}')
+big2bigger_md5sum_0=$(md5sum $B0/${V0}0/big2bigger | awk '{print $1}')
+
+EXPECT $big_md5sum echo $big_md5sum_0
+EXPECT $small_md5sum echo $small_md5sum_0
+EXPECT $big2bigger_md5sum echo $big2bigger_md5sum_0
+EXPECT $bigger2big_md5sum echo $bigger2big_md5sum_0
+
+EXPECT "1" has_holes $B0/${V0}0/big
+EXPECT "1" has_holes $B0/${V0}0/big2bigger
+EXPECT "0" has_holes $B0/${V0}0/bigger2big
+EXPECT "0" has_holes $B0/${V0}0/small
+
+cleanup
diff --git a/tests/volume.rc b/tests/volume.rc
index 9a06687cd86..9e4843e06c4 100644
--- a/tests/volume.rc
+++ b/tests/volume.rc
@@ -300,5 +300,11 @@ function data_written_count {
     echo "$1" | grep "Data Written:$2bytes" | wc -l
 }
 
-
-
+function has_holes {
+        if [ $((`stat -c '%b*%B-%s' -- $1`)) -lt 0 ];
+        then
+                echo "1"
+        else
+                echo "0"
+        fi
+}
diff --git a/xlators/cluster/afr/src/afr-self-heal-data.c b/xlators/cluster/afr/src/afr-self-heal-data.c
index c0385153ff5..c0548d9958f 100644
--- a/xlators/cluster/afr/src/afr-self-heal-data.c
+++ b/xlators/cluster/afr/src/afr-self-heal-data.c
@@ -159,7 +159,7 @@ __afr_selfheal_data_read_write (call_frame_t *frame, xlator_t *this, fd_t *fd,
 		 */
 #define is_last_block(o,b,s) ((s >= o) && (s <= (o + b)))
 		if (HAS_HOLES ((&replies[source].poststat)) &&
-		    offset > replies[i].poststat.ia_size &&
+		    offset >= replies[i].poststat.ia_size &&
 		    !is_last_block (offset, size,
 				    replies[source].poststat.ia_size) &&
 		    (iov_0filled (iovec, count) == 0))
@@ -268,6 +268,31 @@ afr_selfheal_data_restore_time (call_frame_t *frame, xlator_t *this,
 }
 
 static int
+afr_data_self_heal_type_get (afr_private_t *priv, unsigned char *healed_sinks,
+                             int source, struct afr_reply *replies)
+{
+        int type = AFR_SELFHEAL_DATA_FULL;
+        int i = 0;
+
+        if (priv->data_self_heal_algorithm == NULL) {
+                type = AFR_SELFHEAL_DATA_FULL;
+                for (i = 0; i < priv->child_count; i++) {
+                        if (!healed_sinks[i] && i != source)
+                                continue;
+                        if (replies[i].poststat.ia_size) {
+                                type = AFR_SELFHEAL_DATA_DIFF;
+                                break;
+                        }
+                }
+        } else if (strcmp (priv->data_self_heal_algorithm, "full") == 0) {
+                type = AFR_SELFHEAL_DATA_FULL;
+        } else if (strcmp (priv->data_self_heal_algorithm, "diff") == 0) {
+                type = AFR_SELFHEAL_DATA_DIFF;
+        }
+        return type;
+}
+
+static int
 afr_selfheal_data_do (call_frame_t *frame, xlator_t *this, fd_t *fd,
 		      int source, unsigned char *healed_sinks,
 		      struct afr_reply *replies)
@@ -296,14 +321,8 @@ afr_selfheal_data_do (call_frame_t *frame, xlator_t *this, fd_t *fd,
 		"source=%d sinks=%s",
 		uuid_utoa (fd->inode->gfid), source, sinks_str);
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (!healed_sinks[i] && i != source)
-			continue;
-		if (replies[i].poststat.ia_size) {
-			type = AFR_SELFHEAL_DATA_DIFF;
-			break;
-		}
-	}
+        type = afr_data_self_heal_type_get (priv, healed_sinks, source,
+                                            replies);
 
 	iter_frame = afr_copy_frame (frame);
 	if (!iter_frame)