The problem was that the fixes had us generating TLS certs with a
2-day lifetime on the assumption that we'd rotate fairly often. In
fact, we never rotate our TLS keys.
This patch fixes the situation in 2 ways:
1. It bumps the default lifetime back up to one year until we get
rotation in place.
2. It changes tor_tls_context_new() so that it doesn't leak memory
when you call it more than once.
svn:r663
Allow some slop (currently 3 minutes) when checking certificate validity.
Change certificate lifetime from 1 year to 2 days. Since we
regenerate regularly (we regenerate regularly, right??), this
shouldn't be a problem.
Have directories reject descriptors published too far in the future
(currently 30 minutes). If dirservs don't do this:
0) Today is January 1, 2000.
1) A very skewed server publishes descriptor X with a declared
publication time of August 1, 2000.
2) The directory includes X.
3) Because of certificate lifetime issues, nobody can use the
skewed server.
4) The server fixes its skew, and goes to republish a new descriptor Y
with publication time of January 1, 2000.
5) But because the directory already has a "more recent" descriptor X,
it rejects descriptor "Y" as superseded!
This patch should make step 2 go away.
svn:r658
setuid, because after we setuid we don't have the priviledges we
need to setgid anymore, duh. merged switch_user() and
switch_group() into switch_id(), since that code has to be wound
together.
- return -1 from switch_id() if it's not defined to do anything else.
- moved daemoinize(), write_pidfile(), and switch_id() from main.c to
util.c
svn:r656
setuid and setgid respectively, and die if it can't.
(If the User option is set, tor will setgid to the user's gid as well.)
This happens after the pidfile is created, so that in cases where tor
needs to be root to work with the pidfile, it will at least be able to
create it, although it won't be able to delete it. That sucks, but
it's somewhat better than not being able to create the pidfile in the
first place.
svn:r652
so he gets the permissions right.
also this means clients will never need to make the datadirectory.
also remind the admin to fix his clock before setting up his node.
svn:r650
not when we're closing the stream.
this lets us put a payload in the end cell if we want to,
to describe why we're closing the stream.
there are still some places where we don't send the end cell
immediately. i need to track them down. but it's a low priority,
since i've made it send the end cell when we close the stream if
we haven't already sent it.
svn:r640
Roger Dingledine
Nick Mathewson
Paul Syverson
Steven Hazel