Hi Glenn,

On Mon, Aug 18, 2025 at 03:31:59AM -0400, Glenn Strauss wrote:
> On Fri, Aug 15, 2025 at 09:40:39PM +0200, Salvatore Bonaccorso wrote:
> > Hi Glenn,
> > 
> > On Fri, Aug 15, 2025 at 03:19:14AM -0400, Glenn Strauss wrote:
> > > Salvatore: lighttpd is not directly vulnerable to HTTP/2 MadeYouReset
> > > 
> > > As published in https://kb.cert.org/vuls/id/767506
> > > > Vendor Statement
> > > >
> > > > lighttpd is not directly vulnerable to HTTP/2 MadeYouReset. lighttpd 
> > > > tracks request streams with connections to backends, makes a single 
> > > > request on each backend socket connection, and closes the socket (or 
> > > > kill()s the CGI) when the request stream is reset.
> > > 
> > > 
> > > On Fri, Aug 15, 2025 at 06:09:43AM +0200, Salvatore Bonaccorso wrote:
> > > > The following vulnerability was published for lighttpd.
> > > > CVE-2025-8671[0]:
> > > 
> > > ***
> > > Please contact security at lighttpd.net prior to filing a CVE
> > > and prior to publishing a CVE.
> > > ***
> > >
> > > lighttpd 1.4.80 adds *detection* of HTTP/2 MadeYouReset so that log
> > > watchers such as fail2ban can be configured to block offending IPs.
> > 
> > Right, and this is the mentioned references in the bugeport:
> > https://www.lighttpd.net/2025/8/13/1.4.80/
> > https://github.com/lighttpd/lighttpd1.4/commit/8442ca4c699566cdd7369e09690926f403b54fc9
> > 
> > > I see now that the lighttpd release notes could have been more explicit
> > > that lighttpd is not directly vulnerable to MadeYouReset, the same way
> > > that lighttpd was not directly vulnerable to Rapid Reset attacks.
> > > 
> > > While lighttpd 1.4.80 will close connections to offending clients,
> > > an attacker can merely reconnect and continue the attack, so the
> > > disconnection is a small mitigation.  At the end of the day, a DoS
> > > attack is a DoS attack and more effective blocks can be performed
> > > at the firewall or upstream, especially across a farm of independent
> > > lighttpd servers.
> > > 
> > > Adding detection and error logging for independent lighttpd servers
> > > across a server farm is one of the reasons lighttpd 1.4.80 adds
> > > detection and logging for MadeYouReset attacks.
> > 
> > Thanks, this make sense, we will update the status for the
> > security-tracker in Debian.
> > 
> > That is it is surely sensible to make the 1.4.80 update for unstable
> > and so forky and maybe if  feasible add this dedection back as well
> > for trixie and bookworm, what is your take? (this via upcoming point
> > releases).
> > 
> > Regards,
> > Salvatore
> 
> I plan to submit updated packages for lighttpd 1.4.81
> in early Sept before the Trixie point release.
> 
> I will be AFK for a few weeks, and do not want to rush changes if I
> won't be able to respond in a timely manner.
> 
> In the meantime, if Debian Developers would like to backport the
> lighttpd 1.4.80 MadeYouReset detection patch to lighttpd 1.4.79,
> please go ahead:
> https://github.com/lighttpd/lighttpd1.4/commit/8442ca4c699566cdd7369e09690926f403b54fc9
> 
> Otherwise, I plan to do so in September.

Your proposed plan and doing it in semptember with the point release
is perfecly fine.

Thank you,

Regards,
Salvatore

Reply via email to