[Libguestfs] [PATCH] docs: Link to protocol security considerations in uri docs

[Eric Blake]

On Tue, Aug 10, 2021 at 01:08:59PM -0500, Eric Blake wrote:
> Especially useful in light of the recent publishing of
> https://nostarttls.secvuln.info/, which documents a variety of
> implementations vulnerable to downgrade attacks in SMTP and IMAP, as
> well as its caution that that any protocol with a STARTTLS operation
> (which includes NBD) needs to be aware of the potential downgrade
> attacks.
> 
> The NBD protocol documentation already covers what is necessary to
> avoid the effects of a downgrade attack, and all known implementations
> of NBD servers and clients with working NBD_OPT_STARTTLS have at least
> one mode where TLS is mandatory rather than opportunistic.  So I don't
> see this as a CVE against the NBD protocol itself, so much as a worry
> about the potential for future poor implementations that disregard the
> documentation.
> ---
> 
> I'm likely to push this to the NBD spec later this week if it doesn't
> receive any reviews beforehand.

As a followup, I got this reply from Hanno Böck on oss-security:

https://www.openwall.com/lists/oss-security/2021/08/11/8
| The buffering vulnerabilities we found are in STARTTLS implementations
| that have the expectation to enforce a secure connection, but suffer
| from various vulnerabilities in the implementation.

One of the reasons that SMTP and IMAP were particularly problematic in
implementations is that they are line-based protocols, with
arbitrary-sized packets where the length isn't known until the \n is
reached.  Both clients and servers have to choose whether to read one
character at a time (painfully slow) or read ahead into a buffer and
then processing what is in the buffer.  Many of the CVEs raised were
in regards to mishandling such buffers, such as acting on
previously-buffered plaintext even after the switch to encryption.

Thankfully, the NBD protocol has a much more structured handshake
(while different NBD_OPT commands can have different payload lenghts,
at least the header of each packet designates the length in advance,
reducing the need for read-ahead buffering), as well as documentation
that the NBD_OPT_ phase is a lockstep algorithm (neither client nor
server should be building up a buffer of more than one
command/response).

Another aspect of the SMTP/IMAP security holes came from incorrectly
carrying state across the transition to encryption (making a false
assumption that the answer made in plaintext will be the same when
encrypted).  Unfortunately, I have realized that the NBD spec has one
bit of state (NBD_OPT_SET_META_CONTEXT) where it is currently silent
on how to handle that state across a transition to encrypted.  So I
plan on posting a followup patch that matches the actual
implementation of nbdkit in opportunistic mode (qemu-nbd does not
offer opportunistic mode, and nbd-server does not suport
NBD_OPT_SET_META_CONTEXT): a server in opportunistic mode MUST treat
the NBD_OPT_STARTTLS command as wiping out any earlier
NBD_OPT_SET_META_CONTEXT.

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3266
Virtualization:  qemu.org | libvirt.org