+++ This bug was initially created as a clone of Bug #180536 +++ postgresql pgcrypt minor salt generation flaw http://marc.theaimsgroup.com/?l=bugtraq&m=113933370118193&w=2 There is a signedness error in the way postgresql generated the password salts. Worst case scenario is there are only half as many salts available as there should be. This is only an issue if someone can steal the hashed postgresql passwords, and even then it will still take a VERY long time to crack these. I don't know which algorithm we use by default, but if it should take 2 weeks to crack, it will really only take 1 week (probably longer). Part of the problem with fixing this is the already generated passwords are already weak, fixing this will only affect new passwords.. Taking into account it's only database passwords, I'm assigning this "low" Here is the patch. http://cvsweb.openwall.com/cgi/cvsweb.cgi/Owl/packages/glibc/crypt_blowfish/crypt_gensalt.c.diff? r1=1.1;r2=1.2 This issue also affects RHEL3 -- Additional comment from tgl on 2006-02-16 21:01 EST -- This has been fixed upstream in postgresql 7.4.11 (relevant to RHEL4) and 7.3.13 (relevant to RHEL3). There are a fair number of other serious bugs fixed in the upstream updates since we last turned postgresql in RHEL. I think we ought to plan on pushing the updated versions out in the next quarterly update cycle. This particular problem does not seem severe enough to justify an async security erratum, though.
Fixed in postgresql 7.4.13 and 7.3.15, pushed today for other security errata.