XML::LibXML memory leak

[spstansbury]

Greetings!

I have a script that processes the output of a vulnerability scanner. When I have a CVE identifier, I look up the CVSS base metrics from the NVD files (nvd.nist.gov).

Everything works just fine, but as I process thousands of records, I run out of memory, top shows an ever incrementing VSIZE...

The issue is in this block of code, as I can set the $cve_id to skip this else clause and the script trundles along happily.

I know that I am declaring a new "my parser=", etc. everytime, how can I make sure the data structure are deleted/torn down?</p

else {
    # Take the CVE indentifier and get the CVSS vectors
    # parse the CVE identifier to determine what data file to search
    
    my @record_fields = split( /:/, $cve_id );
    $cve_id = $record_fields[1];
    
    $cve_id =~ s/^\s+//;
    $cve_id =~ s/\s+$//;

    my @id_fields = split( /-/, $cve_id );
    my $year = $id_fields[1];

    if ($year < 2003) {
        $data_file = "$nvd_files/nvdcve-2.0-2002.xml";
    }
    else {
        $data_file = "$nvd_files/nvdcve-2.0-" . $year . ".xml";
    }

    # Parse the data file:
    my $cve_parser = XML::LibXML->new();
    my $cve_doc = $cve_parser->parse_file( $data_file );
    my $cve_xc = XML::LibXML::XPathContext->new( $cve_doc->documentEle
+ment() );
    
    # Register the namespaces:
    $cve_xc->registerNs( def => 'http://scap.nist.gov/schema/feed/vuln
+erability/2.0' );
    $cve_xc->registerNs( vuln => 'http://scap.nist.gov/schema/vulnerab
+ility/0.4' );
    $cve_xc->registerNs( cvss => 'http://scap.nist.gov/schema/cvss-v2/
+0.2' );
    
    # Find the appropriate CVE entry in the data source:
    for my $entry ($cve_xc->findnodes("/def:nvd/def:entry[\@id = '$cve
+_id']")) {
        if (my ($metrics) = $cve_xc->findnodes('vuln:cvss/cvss:base_me
+trics', $entry)) {
        
            $av = $cve_xc->find('cvss:access-vector', $metrics);
            $ac = $cve_xc->find('cvss:access-complexity', $metrics);
            $au = $cve_xc->find('cvss:authentication', $metrics);
            $ci = $cve_xc->find('cvss:confidentiality-impact', $metric
+s);
            $ii = $cve_xc->find('cvss:integrity-impact', $metrics);
            $ai = $cve_xc->find('cvss:availability-impact', $metrics);
        }
        else {
            $av = "";
            $ac = "";
            $au = "";
            $ci = "";
            $ii = "";
            $ai = "";
        }
    }
}

As always, thanks for any and all help!

Scott

[ikegami]

The tree can't be freed while you're still using it with $ac, $au, $ci, $ii and $ai.

[Anonymous Monk]

It probably could if you detach $ac/$au/$ci/$ii/$ai, perhaps with

sub XML::LibXML::Node::detach {
    my( $self ) = @);
    $self->parent->removeChild( $self );
}

[ikegami]

There's no reason to go messing with someone else's namespace.

sub detach {
    my( $node ) = @_;
    $node->parentNode->removeChild( $node );
}

detach($node);

would work just as well.

Except it's not enough. It won't separate it from the document.

$ perl -MXML::LibXML -E'
   my $node;
   {
      my $xml = "<root><foo><bar/></foo></root>";
      my $doc = XML::LibXML->new->parse_string($xml);
      ($node) = $doc->findnodes("//bar");
      $node->parentNode->removeChild($node);
   }
   {
      my $doc = $node->ownerDocument;
      say "owner=", $doc;
      if ($doc) { say $_->nodeName for $doc->findnodes("//*"); }
   }
'
owner=XML::LibXML::Document=SCALAR(0x817bcb8)
root
foo

You need to give the node a new document.

$ perl -MXML::LibXML -E'
   my $foster_home = XML::LibXML::Document->new("1.0", "UTF-8");
   my $node;
   {
      my $xml = "<root><foo><bar/></foo></root>";
      my $doc = XML::LibXML->new->parse_string($xml);
      ($node) = $doc->findnodes("//bar");
      $node->setOwnerDocument($foster_home);
   }
   {
      my $doc = $node->ownerDocument;
      say "owner=", $doc;
      if ($doc) { say $_->nodeName for $doc->findnodes("//*"); }
   }
'
owner=XML::LibXML::Document=SCALAR(0x832af38)

Note that transfers the node's children too.

[Jenda]

Use XML::Twig or XML::Rules and process the file in chunks. By the time your loop runs it'a already too late and the memory has already been wasted.