From: Xuehan Xu <xuxuehan@qianxin.com>
Date: Sun, 8 Sep 2024 05:27:07 +0000 (+0800)
Subject: crimson/osd/pg: rollback ops by copying obc beforehand and recover after
X-Git-Tag: v20.0.0~1041^2~3
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=fc41513d4e990c3a7cba38f100e7e6275dfdf849;p=ceph.git

crimson/osd/pg: rollback ops by copying obc beforehand and recover after
errors

Fixes: https://tracker.ceph.com/issues/63844
Signed-off-by: Xuehan Xu <xuxuehan@qianxin.com>
---

diff --git a/src/crimson/osd/object_context.h b/src/crimson/osd/object_context.h
index 4148e3b592c0..e17af91e3ade 100644
--- a/src/crimson/osd/object_context.h
+++ b/src/crimson/osd/object_context.h
@@ -76,6 +76,11 @@ public:
   ObjectContext(hobject_t hoid) : lock(hoid),
                                   obs(std::move(hoid)) {}
 
+  void update_from(const ObjectContext &obc) {
+    obs = obc.obs;
+    ssc = obc.ssc;
+  }
+
   const hobject_t &get_oid() const {
     return obs.oi.soid;
   }
diff --git a/src/crimson/osd/ops_executer.h b/src/crimson/osd/ops_executer.h
index 0448be407d85..623e3d90d56a 100644
--- a/src/crimson/osd/ops_executer.h
+++ b/src/crimson/osd/ops_executer.h
@@ -619,7 +619,10 @@ OpsExecuter::RollbackHelper<Func>::rollback_obc_if_modified(
     ox->obc->get_oid(),
     e,
     need_rollback);
-  return need_rollback ? func(*ox->obc) : interruptor::now();
+  if (need_rollback) {
+    func(ox->obc);
+  }
+  return interruptor::now();
 }
 
 // PgOpsExecuter -- a class for executing ops targeting a certain PG.
diff --git a/src/crimson/osd/pg.cc b/src/crimson/osd/pg.cc
index b4a0074b382f..6038dd655a40 100644
--- a/src/crimson/osd/pg.cc
+++ b/src/crimson/osd/pg.cc
@@ -963,6 +963,15 @@ PG::BackgroundProcessLock::lock() noexcept
   return interruptor::make_interruptible(mutex.lock());
 }
 
+// We may need to rollback the ObjectContext on failed op execution.
+// Copy the current obc before mutating it in order to recover on failures.
+ObjectContextRef duplicate_obc(const ObjectContextRef &obc) {
+  ObjectContextRef object_context = new ObjectContext(obc->obs.oi.soid);
+  object_context->obs = obc->obs;
+  object_context->ssc = new SnapSetContext(*obc->ssc);
+  return object_context;
+}
+
 template <class Ret, class SuccessFunc, class FailureFunc>
 PG::do_osd_ops_iertr::future<PG::pg_rep_op_fut_t<Ret>>
 PG::do_osd_ops_execute(
@@ -975,9 +984,9 @@ PG::do_osd_ops_execute(
   FailureFunc&& failure_func)
 {
   assert(ox);
-  auto rollbacker = ox->create_rollbacker([this] (auto& obc) {
-    return obc_loader.reload_obc(obc).handle_error_interruptible(
-      load_obc_ertr::assert_all{"can't live with object state messed up"});
+  auto rollbacker = ox->create_rollbacker(
+    [object_context=duplicate_obc(obc)] (auto& obc) mutable {
+    obc->update_from(*object_context);
   });
   auto failure_func_ptr = seastar::make_lw_shared(std::move(failure_func));
   return interruptor::do_for_each(ops, [ox](OSDOp& osd_op) {