test/librbd: Add a stress test that reproduces a crash during discard journaling

See the comments in DiscardWithPruneWriteOverlap for details.

Signed-off-by: Joshua Baergen <jbaergen@digitalocean.com>
(cherry picked from commit e8d54e3c9faeddedc2890294556cd66095b83be4)

diff --git a/src/test/librbd/CMakeLists.txt b/src/test/librbd/CMakeLists.txt
index 6d8371fd22f68d5fb042e121f8409a22db2dda73..6fdba5a6c058187bce0653da1c81094440123ce9 100644 (file)
--- a/src/test/librbd/CMakeLists.txt
+++ b/src/test/librbd/CMakeLists.txt
@@ -20,7 +20,8 @@ set(librbd_test
   test_Operations.cc
   test_Trash.cc
   journal/test_Entries.cc
-  journal/test_Replay.cc)
+  journal/test_Replay.cc
+  journal/test_Stress.cc)
 add_library(rbd_test STATIC ${librbd_test})
 target_link_libraries(rbd_test PRIVATE
   rbd_test_support

diff --git a/src/test/librbd/journal/test_Stress.cc b/src/test/librbd/journal/test_Stress.cc
new file mode 100644 (file)
index 0000000..752ecf0
--- /dev/null
+++ b/src/test/librbd/journal/test_Stress.cc
@@ -0,0 +1,101 @@
+// -*- mode:C++; tab-width:8; c-basic-offset:2; indent-tabs-mode:t -*-
+// vim: ts=8 sw=2 smarttab
+
+#include "test/librbd/test_fixture.h"
+#include "test/librbd/test_support.h"
+#include "cls/rbd/cls_rbd_types.h"
+#include "cls/journal/cls_journal_types.h"
+#include "cls/journal/cls_journal_client.h"
+#include "journal/Journaler.h"
+#include "librbd/ExclusiveLock.h"
+#include "librbd/ImageCtx.h"
+#include "librbd/ImageState.h"
+#include "librbd/ImageWatcher.h"
+#include "librbd/internal.h"
+#include "librbd/Journal.h"
+#include "librbd/Operations.h"
+#include "librbd/api/Io.h"
+#include "librbd/api/Snapshot.h"
+#include "librbd/io/AioCompletion.h"
+#include "librbd/io/ImageDispatchSpec.h"
+#include "librbd/io/ImageRequest.h"
+#include "librbd/io/ReadResult.h"
+#include "librbd/journal/Types.h"
+
+void register_test_journal_stress() {
+}
+
+namespace librbd {
+namespace journal {
+
+class TestJournalStress : public TestFixture {
+};
+
+TEST_F(TestJournalStress, DiscardWithPruneWriteOverlap) {
+  REQUIRE_FEATURE(RBD_FEATURE_JOURNALING);
+
+  // Overlap discards and writes while discard pruning is occurring. This tests
+  // the conditions under which https://tracker.ceph.com/issues/63422 occurred.
+
+  // Create an image that is multiple objects so that we can force multiple
+  // image extents on the discard path.
+  CephContext* cct = reinterpret_cast<CephContext*>(_rados.cct());
+  auto object_size = 1ull << cct->_conf.get_val<uint64_t>("rbd_default_order");
+  auto image_size = 4 * object_size;
+
+  // Write-around cache required for overlapping I/O delays.
+  cct->_conf.set_val_or_die("rbd_cache_writethrough_until_flush", "false");
+  cct->_conf.set_val_or_die("rbd_cache_policy", "writearound");
+
+  auto image_name = get_temp_image_name();
+  ASSERT_EQ(0, create_image_pp(m_rbd, m_ioctx, image_name, image_size));
+
+  librbd::ImageCtx *ictx;
+  ASSERT_EQ(0, open_image(image_name, &ictx));
+
+  std::thread write_thread(
+    [ictx, object_size]() {
+      std::string payload(object_size, '1');
+
+      for (auto i = 0; i < 200; i++) {
+        // Alternate overlaps with the two objects that the discard below
+        // touches.
+        for (auto offset = object_size;
+             offset < object_size * 3;
+             offset += object_size) {
+          bufferlist payload_bl;
+          payload_bl.append(payload);
+          auto aio_comp = new librbd::io::AioCompletion();
+          api::Io<>::aio_write(*ictx, aio_comp, 0, payload.size(),
+                               std::move(payload_bl), 0, true);
+          ASSERT_EQ(0, aio_comp->wait_for_complete());
+          aio_comp->release();
+        }
+      }
+    }
+  );
+
+  auto discard_exit = false;
+  std::thread discard_thread(
+    [ictx, object_size, &discard_exit]() {
+      while (!discard_exit) {
+        // We offset the discard by -4096 bytes and set discard granularity to
+        // 8192; this should cause two image extents to be formed in
+        // AbstractImageWriteRequest<I>::send_request() on objects 1 and 2,
+        // overlapping with the writes above.
+        auto aio_comp = new librbd::io::AioCompletion();
+        api::Io<>::aio_discard(*ictx, aio_comp, object_size - 4096,
+                               2 * object_size, 8192, true);
+        ASSERT_EQ(0, aio_comp->wait_for_complete());
+        aio_comp->release();
+      }
+    }
+  );
+
+  write_thread.join();
+  discard_exit = true;
+  discard_thread.join();
+}
+
+} // namespace journal
+} // namespace librbd

diff --git a/src/test/librbd/test_main.cc b/src/test/librbd/test_main.cc
index 2ff9f69dea97731bc760ef0e841a973c5297bc21..82b72b1ef7e85e424764faa89b8043e2ad9f958b 100644 (file)
--- a/src/test/librbd/test_main.cc
+++ b/src/test/librbd/test_main.cc
@@ -17,6 +17,7 @@ extern void register_test_image_watcher();
 extern void register_test_internal();
 extern void register_test_journal_entries();
 extern void register_test_journal_replay();
+extern void register_test_journal_stress();
 extern void register_test_migration();
 extern void register_test_mirroring();
 extern void register_test_mirroring_watcher();
@@ -37,6 +38,7 @@ int main(int argc, char **argv)
   register_test_internal();
   register_test_journal_entries();
   register_test_journal_replay();
+  register_test_journal_stress();
   register_test_migration();
   register_test_mirroring();
   register_test_mirroring_watcher();