osd: fix backoff vs reset race

In OSD::ms_handle_reset, we clear session->con before removing any
backoffs.  That means we have to check if con has been cleared after any
call to have_backoff, lest we race with ms_handle_reset and it removes the
backoffs but we don't realize our client session is disconnected.

Introduce a helper to do both these checks in a safe way, simplifying
callers while we're at it.

Signed-off-by: Sage Weil <sage@redhat.com>

1 files changed, 20 insertions, 0 deletions

diff --git a/src/osd/Session.cc b/src/osd/Session.cc
index abe1a0c9eba..2de44271344 100644
--- a/src/osd/Session.cc
+++ b/src/osd/Session.cc
@@ -81,3 +81,23 @@ void Session::ack_backoff(
   }
   assert(!backoff_count == backoffs.empty());
 }
+
+bool Session::check_backoff(
+  CephContext *cct, spg_t pgid, const hobject_t& oid, Message *m)
+{
+  BackoffRef b(have_backoff(pgid, oid));
+  if (b) {
+    dout(10) << __func__ << " session " << this << " has backoff " << *b
+	     << " for " << *m << dendl;
+    assert(!b->is_acked() || !g_conf->osd_debug_crash_on_ignored_backoff);
+    return true;
+  }
+  // we may race with ms_handle_reset.  it clears session->con before removing
+  // backoffs, so if we see con is cleared here we have to abort this
+  // request.
+  if (!con) {
+    dout(10) << __func__ << " session " << this << " disconnected" << dendl;
+    return true;
+  }
+  return false;
+}