Merge PR ceph#61173 into main

* refs/pull/61173/head:

Reviewed-by: Venky Shankar <[email protected]>
Reviewed-by: Milind Changire <[email protected]>
Reviewed-by: Anthony D Atri <[email protected]>

Author: vshankar

doc/cephfs/scrub.rst

@@ -150,6 +150,8 @@ These above named MDS damages can be repaired by using the following command::
 If scrub is able to repair the damage, the corresponding entry is automatically
 removed from the damage table.
 
+Note: A scrub invoked with the ``repair`` option can identify an damaged hard link but not repair it.
+
 
 Evaluate strays using recursive scrub
 =====================================

qa/suites/fs/functional/tasks/scrub.yaml

@@ -8,6 +8,7 @@ overrides:
       - bad backtrace on inode
       - overall HEALTH_
       - \(MDS_TRIM\)
+      - object missing on disk
     conf:
       mds:
         mds log max segments: 1

qa/tasks/cephfs/test_scrub_checks.py

@@ -477,6 +477,44 @@ def test_scrub_split_dirfrags(self):
             timeout=30
         )
 
+    def test_scrub_remote_link(self):
+        """
+        test scrub remote link
+        """
+        test_dir_path = "test_dir"
+        self.mount_a.run_shell(["mkdir", test_dir_path])
+        file_dir_path = os.path.join(test_dir_path, "file_dir")
+        self.mount_a.run_shell(["mkdir", file_dir_path])
+        file_path = os.path.join(file_dir_path, "test_file.txt")
+        link_path = os.path.join(test_dir_path, "test_link")
+        abs_link_path = "/" + link_path
+        self.mount_a.run_shell(["touch", file_path])
+        self.mount_a.run_shell(["ln", file_path, link_path])
+        file_ino = self.mount_a.path_to_ino(file_path)
+        dir_ino = self.mount_a.path_to_ino(file_dir_path)
+        rados_obj_file = "{ino:x}.00000000".format(ino=file_ino)
+        rados_obj_dir = "{ino:x}.00000000".format(ino=dir_ino)
+        self.fs.flush()
+        self.fs.rados(["rm", rados_obj_file], pool=self.fs.get_data_pool_name())
+        self.fs.rados(["rm", rados_obj_dir], pool=self.fs.get_metadata_pool_name())
+        self.fs.mds_fail_restart()
+        self.fs.wait_for_daemons()
+        status = self.fs.mds_asok(['status'])
+        self.assertEqual("up:active", str(status['state']))
+        mds_rank = self.fs.get_rank()['rank']
+        success_validator = lambda j, r: self.json_validator(j, r, "return_code", 0)
+        scrub_json = self.tell_command(mds_rank, 
+            "scrub start /{0} recursive force".format(test_dir_path), 
+                success_validator)
+        self.assertEqual(
+            self.fs.wait_until_scrub_complete(tag=scrub_json["scrub_tag"]), True)
+        damage_json = self.tell_command(mds_rank, "damage ls")
+        found_remote_link_damage = False
+        for entry in damage_json:
+            if entry["path"] == abs_link_path :
+                found_remote_link_damage = True
+        self.assertEqual(found_remote_link_damage, True)
+
     def test_stray_evaluation_with_scrub(self):
         """
         test that scrub can iterate over ~mdsdir and evaluate strays
@@ -501,7 +539,7 @@ def json_validator(json_out, rc, element, expected_value):
                 jv=element_value, ev=expected_value)
         return True, "Succeeded"
 
-    def tell_command(self, mds_rank, command, validator):
+    def tell_command(self, mds_rank, command, validator=None):
         log.info("Running command '{command}'".format(command=command))
 
         command_list = command.split()
@@ -510,9 +548,10 @@ def tell_command(self, mds_rank, command, validator):
         log.info("command '{command}' returned '{jout}'".format(
                      command=command, jout=jout))
 
-        success, errstring = validator(jout, 0)
-        if not success:
-            raise AsokCommandFailedError(command, 0, jout, errstring)
+        if validator:
+            success, errstring = validator(jout, 0)
+            if not success:
+                raise AsokCommandFailedError(command, 0, jout, errstring)
         return jout
 
     @staticmethod

src/mds/ScrubStack.cc

@@ -487,6 +487,19 @@ void ScrubStack::scrub_dir_inode_final(CInode *in)
   return;
 }
 
+void ScrubStack::identify_remote_link_damage(CDentry *dn) {
+
+  CDentry::linkage_t *dnl = dn->get_linkage();
+  CInode *remote_inode = mdcache->get_inode(dnl->get_remote_ino());
+  if (!remote_inode) {
+    if (mdcache->mds->damage_table.is_remote_damaged(dnl->get_remote_ino())) {
+      dout(4) << "scrub: remote dentry points to damaged ino " << *dn << dendl;
+      return;
+    }
+    mdcache->open_remote_dentry(dn, true, new C_MDSInternalNoop());
+  }
+}
+
 void ScrubStack::scrub_dirfrag(CDir *dir, bool *done)
 {
   ceph_assert(dir != NULL);
@@ -530,7 +543,7 @@ void ScrubStack::scrub_dirfrag(CDir *dir, bool *done)
       if (dnl->is_primary()) {
 	_enqueue(dnl->get_inode(), header, false);
       } else if (dnl->is_remote() || dnl->is_referent_remote()) {
-	// TODO: check remote linkage
+        identify_remote_link_damage(dn);
       }
     }
   }

src/mds/ScrubStack.h

@@ -212,6 +212,12 @@ class ScrubStack {
    */
   void scrub_file_inode(CInode *in);
 
+  /**
+   * Scrub a file inode.
+   * @param dn The remote dentry to identify
+   */
+  void identify_remote_link_damage(CDentry *dn);
+
   /**
    * Callback from completion of CInode::validate_disk_state
    * @param in The inode we were validating