A flaw was found in the way that curl sanitized URLs.  The upstream advisory [1] reports:

   libcurl is vulnerable to a data injection attack for certain protocols
   through control characters embedded or percent-encoded in URLs.

   When parsing URLs, libcurl's parser is very laxed and liberal and only
   parses as little as possible and lets as much as possible through as long as
   it can figure out what to do.

   In the specific process when libcurl extracts the file path part from a
   given URL, it didn't always verify the data or escape control characters
   properly before it passed the file path on to the protocol-specific code
   that then would use it for its protocol business.

   This passing through of control characters could be exploited by someone who
   would be able to pass in a handicrafted URL to libcurl. Lots of libcurl
   using applications let users enter URLs in one form or another and not all
   of these check the input carefully to prevent malicious ones.

   A malicious user might pass in %0d%0a to get treated as CR LF by libcurl,
   and by using this fact a user can trick for example a POP3 client to delete
   a message instead of getting it or trick an SMTP server to send an
   unintended message.

   This vulnerability can be used to fool libcurl with the following protocols:
   IMAP, POP3 and SMTP.
                                                                                                                                                                                    There is no known exploit for this problem.


This flaw only affects curl versions 7.20.0 up to and including 7.23.1  It is corrected in 7.24.0 by scanning for a range of "bad codes" in the path part of URLs so that they are rejected before any protocol code even can consider using them.

This flaw has been assigned the name CVE-2012-0036.

[1] http://curl.haxx.se/docs/security.html

Statement:

Not vulnerable. This issue did not affect the versions of curl as shipped with Red Hat Enterprise Linux 4, 5 or 6.