Thanks for the reply, yes, you are right, it matches actually. The regular expression has no problem to match binary data. My problem turns out to be in XML::Simple->XMLin.

use XML::Simple;
my $ref=XMLin( "xml_invalid_char.input" );
print Dumper $ref;
[download]

Here is the input file content:

<description>?s«ndjfr?1 S334µµ!</description>
[download]

(Some strange char can not past here) The error is: not well-formed (invalid token) at line 1, column 22, byte 23 at /usr/lib/perl5/XML/Parser.pm line 187 Is there a way to let XMLin take in the input?