From: Milind Changire <mchangir@redhat.com>
Date: Tue, 26 Sep 2023 10:50:50 +0000 (+0530)
Subject: mds/scrub: enqueue all child frags for a given fragset
X-Git-Tag: testing/wip-batrick-testing-20240411.154038~648^2
X-Git-Url: http://git.apps.os.sepia.ceph.com/?a=commitdiff_plain;h=c20b4d706dfe32e08c5301676fd9144b052a4347;p=ceph-ci.git

mds/scrub: enqueue all child frags for a given fragset

Problem:
fragsets sent over to replicas for scrubbing are simplified i.e. they
are bit representation of the lease common ancestors of the frags that
need scrubbing on that replica. A search operation of a frag in the
frasget often fails to match exactly with the frags delegated to the
replica causing the scrub item to infinitely be held in the scrub stack.

Solution:
Test if the frag in the fragset sent over to the replica contains the
delegated frag as a child to accept it for scrubbing.

Fixes: https://tracker.ceph.com/issues/62658
Signed-off-by: Milind Changire <mchangir@redhat.com>
---

diff --git a/src/mds/ScrubStack.cc b/src/mds/ScrubStack.cc
index 6d799343f14..047bf3ba822 100644
--- a/src/mds/ScrubStack.cc
+++ b/src/mds/ScrubStack.cc
@@ -892,22 +892,30 @@ void ScrubStack::handle_scrub(const cref_t<MMDSScrub> &m)
 
       std::vector<CDir*> dfs;
       MDSGatherBuilder gather(g_ceph_context);
+      frag_vec_t frags;
+      diri->dirfragtree.get_leaves(frags);
       for (const auto& fg : m->get_frags()) {
-	CDir *dir = diri->get_dirfrag(fg);
-	if (!dir) {
-	  dout(10) << __func__ << " no frag " << fg << dendl;
-	  continue;
-	}
-	if (!dir->is_auth()) {
-	  dout(10) << __func__ << " not auth " << *dir << dendl;
-	  continue;
-	}
-	if (!dir->can_auth_pin()) {
-	  dout(10) << __func__ << " can't auth pin " << *dir <<  dendl;
-	  dir->add_waiter(CDir::WAIT_UNFREEZE, gather.new_sub());
-	  continue;
+	for (auto f : frags) {
+	  if (!fg.contains(f)) {
+	    dout(20) << __func__ << " skipping " << f << dendl;
+	    continue;
+	  }
+	  CDir *dir = diri->get_or_open_dirfrag(mdcache, f);
+	  if (!dir) {
+	    dout(10) << __func__ << " no frag " << f << dendl;
+	    continue;
+	  }
+	  if (!dir->is_auth()) {
+	    dout(10) << __func__ << " not auth " << *dir << dendl;
+	    continue;
+	  }
+	  if (!dir->can_auth_pin()) {
+	    dout(10) << __func__ << " can't auth pin " << *dir <<  dendl;
+	    dir->add_waiter(CDir::WAIT_UNFREEZE, gather.new_sub());
+	    continue;
+	  }
+	  dfs.push_back(dir);
 	}
-	dfs.push_back(dir);
       }
 
       if (gather.has_subs()) {