cluster/ec: honor contention notifications for partially acquired locks

EC was ignoring lock contention notifications received while a lock was
being acquired. When a lock is partially acquired (some bricks have
granted the lock but some others not yet) we can receive notifications
from acquired bricks, which should be honored, since we may not receive
more notifications after that.

Since EC was ignoring them, once the lock was acquired, it was not
released until the eager-lock timeout, causing unnecessary delays on
other clients.

This fix takes into consideration the notifications received before
having completed the full lock acquisition. After that, the lock will
be releaed as soon as possible.

Fixes: bz#1708156
Change-Id: I2a306dbdb29fb557dcab7788a258bd75d826cc12
Signed-off-by: Xavi Hernandez <xhernandez@redhat.com>

2 files changed, 55 insertions, 1 deletions

diff --git a/tests/bugs/ec/bug-1708156-honor-inodelk-contention-notify-on-partial-locks.t b/tests/bugs/ec/bug-1708156-honor-inodelk-contention-notify-on-partial-locks.t
new file mode 100644
index 00000000000..67fdb184b46
--- /dev/null
+++ b/tests/bugs/ec/bug-1708156-honor-inodelk-contention-notify-on-partial-locks.t
@@ -0,0 +1,54 @@
+#!/bin/bash
+
+. $(dirname $0)/../../include.rc
+. $(dirname $0)/../../volume.rc
+
+function do_ls() {
+    local dir="${1}"
+    local i
+
+    for i in {1..50}; do
+        ls -l $M0/${dir} >/dev/null &
+        ls -l $M1/${dir} >/dev/null &
+        ls -l $M2/${dir} >/dev/null &
+        ls -l $M3/${dir} >/dev/null &
+    done
+    wait
+}
+
+function measure_time() {
+    {
+        LC_ALL=C
+        time -p "${@}"
+    } 2>&1 | awk '/^real/ { print $2 * 1000 }'
+}
+
+cleanup
+
+TEST glusterd
+TEST pidof glusterd
+TEST $CLI volume create $V0 disperse 6 redundancy 2 $H0:$B0/${V0}{0..5}
+
+TEST $CLI volume set $V0 disperse.eager-lock on
+TEST $CLI volume set $V0 disperse.other-eager-lock on
+TEST $CLI volume set $V0 features.locks-notify-contention on
+TEST $CLI volume set $V0 disperse.eager-lock-timeout 10
+TEST $CLI volume set $V0 disperse.other-eager-lock-timeout 10
+
+TEST $CLI volume start $V0
+TEST $GFS --volfile-id=/$V0 --volfile-server=$H0 $M0
+TEST $GFS --volfile-id=/$V0 --volfile-server=$H0 $M1
+TEST $GFS --volfile-id=/$V0 --volfile-server=$H0 $M2
+TEST $GFS --volfile-id=/$V0 --volfile-server=$H0 $M3
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "6" ec_child_up_count $V0 0 $M0
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "6" ec_child_up_count $V0 0 $M1
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "6" ec_child_up_count $V0 0 $M2
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "6" ec_child_up_count $V0 0 $M3
+TEST mkdir $M0/dir
+TEST touch $M0/dir/file.{1..10}
+
+# Run multiple 'ls' concurrently from multiple clients so that they collide and
+# cause partial locks.
+TEST [[ $(measure_time do_ls dir) -lt 10000 ]]
+
+cleanup
diff --git a/xlators/cluster/ec/src/ec-common.c b/xlators/cluster/ec/src/ec-common.c
index b1ba5e96d3e..e85aa8bf43f 100644
--- a/xlators/cluster/ec/src/ec-common.c
+++ b/xlators/cluster/ec/src/ec-common.c
@@ -2497,7 +2497,7 @@ ec_lock_release(ec_t *ec, inode_t *inode)
         goto done;
     }
     lock = ctx->inode_lock;
-    if ((lock == NULL) || !lock->acquired || lock->release) {
+    if ((lock == NULL) || lock->release) {
         goto done;
     }