test/librbd: actually alternate overlaps in DiscardWithPruneWriteOverlap

Make sense of the inner loop in write_thread.  The crash on
"it != m_events.end()" assert reproduces even faster this way.

Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
(cherry picked from commit d790b57b4bd3868cf12941c565b6906534e19f91)

diff --git a/src/test/librbd/journal/test_Stress.cc b/src/test/librbd/journal/test_Stress.cc
index 5abdc8bae80715d7f358aa910ccae92f776af301..d3df9147ae6aedeefa335352beb924dbb92f8035 100644 (file)
--- a/src/test/librbd/journal/test_Stress.cc
+++ b/src/test/librbd/journal/test_Stress.cc
@@ -86,7 +86,7 @@ TEST_F(TestJournalStress, DiscardWithPruneWriteOverlap) {
           bufferlist payload_bl;
           payload_bl.append(payload);
           auto aio_comp = new librbd::io::AioCompletion();
-          api::Io<>::aio_write(*ictx, aio_comp, 0, payload.size(),
+          api::Io<>::aio_write(*ictx, aio_comp, offset, payload.size(),
                                std::move(payload_bl), 0, true);
           ASSERT_EQ(0, aio_comp->wait_for_complete());
           aio_comp->release();