Um 23:02 Uhr am 30.01.06 schrieb Florian Weimer: > Sven Hartge: >> Um 15:15 Uhr am 30.01.06 schrieb Florian Weimer:
>>> Turns out the patch was broken. This one should be better. The >>> comments above still apply. >> Sorry, but I patched and recompiled the exim4-package from Sarge, but >> any encrypted mail transfer nearly empties the entropy pool. > Again, this is expected -- I tried to fix the blocking problem, not the > entropy consumption as such. The entropy consumption is really a > libgcrypt issue; it does not make much sense to work around it in each > application individually. Let me sum up all the pieces to see, if I get everything in the right way. a) Exim uses GnuTLS in a way, which causes it to use /dev/random to aquire strictly random bytes on every encrypted connection b) GnuTLS uses much more random bytes to initialize itself than OpenSSL. A combination of A and B leads to a possible hole for a DoS attack, if I am able to drain the entropy pool, because of A, Exim will block until enough entropy is regathered. While fixing A is generally a Good Idea (in my opinion), so Exim does not block if there is no entropy available (because of whatever reason). But the real winner would be a fix to B so that the entropy pool does not get drained so fast and would not only benefit Exim but any other program using GnuTLS. Is this summary correct so far? Grüße, Sven. -- Sven Hartge -- professioneller Unix-Geek Meine Gedanken im Netz: http://www.svenhartge.de/ Achtung, neue Mail-Adresse: [EMAIL PROTECTED]