According to the manpage, bridges use P256 for conformity and relays
use P224 for speed. But skruffy points out that we've gotten it
backwards in the code.
In this patch, we make the default P256 for everybody.
Fixes bug 9780; bugfix on 0.2.4.8-alpha.
To fix#6033, we disabled TLS 1.1 and 1.2. Eventually, OpenSSL fixed
the bug behind #6033.
I've considered alternate implementations that do more testing to see
if there's secretly an OpenSSL 1.0.1c or something that secretly has a
backport of the OpenSSL 1.0.1e fix, and decided against it on the
grounds of complexity.
We previously used FILENAME_PRIVATE identifiers mostly for
identifiers exposed only to the unit tests... but also for
identifiers exposed to the benchmarker, and sometimes for
identifiers exposed to a similar module, and occasionally for no
really good reason at all.
Now, we use FILENAME_PRIVATE identifiers for identifiers shared by
Tor and the unit tests. They should be defined static when we
aren't building the unit test, and globally visible otherwise. (The
STATIC macro will keep us honest here.)
For identifiers used only by the unit tests and never by Tor at all,
on the other hand, we wrap them in #ifdef TOR_UNIT_TESTS.
This is not the motivating use case for the split test/non-test
build system; it's just a test example to see how it works, and to
take a chance to clean up the code a little.
This is meant to avoid conflict with the built-in log() function in
math.h. It resolves ticket 7599. First reported by dhill.
This was generated with the following perl script:
#!/usr/bin/perl -w -i -p
s/\blog\(LOG_(ERR|WARN|NOTICE|INFO|DEBUG)\s*,\s*/log_\L$1\(/g;
s/\blog\(/tor_log\(/g;
This implements the server-side of proposal 198 by detecting when
clients lack the magic list of ciphersuites that indicates that
they're lying faking some ciphers they don't really have. When
clients lack this list, we can choose any cipher that we'd actually
like. The newly allowed ciphersuites are, currently, "All ECDHE-RSA
ciphers that openssl supports, except for ECDHE-RSA-RC4".
The code to detect the cipher list relies on on (ab)use of
SSL_set_session_secret_cb.
We already use this classification for deciding whether (as a server)
to do a v2/v3 handshake, and we're about to start using it for
deciding whether we can use good ciphersuites too.
This is less easy than you might think; we can't just look at the
client ciphers list, since openssl doesn't remember client ciphers if
it doesn't know about them. So we have to keep a list of the "v2"
ciphers, with the ones we don't know about removed.
Apparently some compilers like to eliminate memset() operations on
data that's about to go out-of-scope. I've gone with the safest
possible replacement, which might be a bit slow. I don't think this
is critical path in any way that will affect performance, but if it
is, we can work on that in 0.2.4.
Fixes bug 7352.
OpenSSL 1.0.0 added an implementation of TLS session tickets, a
"feature" that let session resumption occur without server-side state
by giving clients an encrypted "ticket" that the client could present
later to get the session going again with the same keys as before.
OpenSSL was giving the keys to decrypt these tickets the lifetime of
the SSL contexts, which would have been terrible for PFS if we had
long-lived SSL contexts. Fortunately, we don't. Still, it's pretty
bad. We should also drop these, since our use of the extension stands
out with our non-use of session cacheing.
Found by nextgens. Bugfix on all versions of Tor when built with
openssl 1.0.0 or later. Fixes bug 7139.
097 hasn't seen a new version since 2007; we can drop support too.
This lets us remove our built-in sha256 implementation, and some
checks for old bugs.
This is a feature removal: we no longer fake any ciphersuite other
than the not-really-standard SSL_RSA_FIPS_WITH_3DES_EDE_CBC_SHA
(0xfeff). This change will let servers rely on our actually
supporting what we claim to support, and thereby let Tor migrate to
better TLS ciphersuites.
As a drawback, Tor instances that use old openssl versions and
openssl builds with ciphers disabled will no longer give the
"firefox" cipher list.
Also, try to resolve some doxygen issues. First, define a magic
"This is doxygen!" macro so that we take the correct branch in
various #if/#else/#endifs in order to get the right documentation.
Second, add in a few grouping @{ and @} entries in order to get some
variables and fields to get grouped together.
It appears that when OpenSSL negotiates a 1.1 or 1.2 connection, and it
decides to renegotiate, the client will send a record with version "1.0"
rather than with the current TLS version. This would cause the
connection to fail whenever both sides had OpenSSL 1.0.1, and the v2 Tor
handshake was in use.
As a workaround, disable TLS 1.1 and TLS 1.2. When a later version of
OpenSSL is released, we can make this conditional on running a fixed
version of OpenSSL.
Alternatively, we could disable TLS 1.1 and TLS 1.2 only on the client
side. But doing it this way for now means that we not only fix TLS with
patched clients; we also fix TLS when the server has this patch and the
client does not. That could be important to keep the network running
well.
Fixes bug 6033.
We've been only treating SW_SERVER_HELLO_A as meaning that an SSL
handshake was happening. But that's not right: if the initial
attempt to write a ServerHello fails, we would get a callback in
state SW_SERVER_HELLO_B instead.
(That's "instead" and not "in addition": any failed attempt to write
the hello will fail and cause the info callback not to get written.)
Fix for bug 4592; bugfix on 0.2.0.13-alpha.
This tells the windows headers to give us definitions that didn't
exist before XP -- like the ones that we need for IPv6 support.
See bug #5861. We didn't run into this issue with mingw, since
mingw doesn't respect _WIN32_WINNT as well as it should for some of
its definitions.
If the client uses a v2 cipherlist on the renegotiation handshake,
it looks as if they could fail to get a good cert chain from the
server, since they server would re-disable certificate chaining.
This patch makes it so the code that make the server side of the
first v2 handshake special can get called only once.
Fix for 4591; bugfix on 0.2.0.20-rc.
This commit is completely mechanical; I used this perl script to make it:
#!/usr/bin/perl -w -i.bak -p
if (/^\s*\#/) {
s/MS_WINDOWS/_WIN32/g;
s/\bWIN32\b/_WIN32/g;
}
- Rename tor_tls_got_server_hello() to tor_tls_got_client_hello().
- Replaced some aggressive asserts with LD_BUG logging.
They were the innocent "I believe I understand how these callbacks
work, and this assert proves it" type of callbacks, and not the "If
this statement is not true, computer is exploding." type of
callbacks.
- Added a changes file.
* Make tor_tls_context_new internal to tortls.c, and return the new
tor_tls_context_t from it.
* Add a public tor_tls_context_init wrapper function to replace it.
Conflicts:
src/or/main.c
src/or/router.c
SSL_read(), SSL_write() and SSL_do_handshake() can always progress the
SSL protocol instead of their normal operation, this means that we
must be checking for needless renegotiations after they return.
Introduce tor_tls_got_excess_renegotiations() which makes the
tls->server_handshake_count > 2
check for us, and use it in tor_tls_read() and tor_tls_write().
Cases that should not be handled:
* SSL_do_handshake() is only called by tor_tls_renegotiate() which is a
client-only function.
* The SSL_read() in tor_tls_shutdown() does not need to be handled,
since SSL_shutdown() will be called if SSL_read() returns an error.
Since we check for naughty renegotiations using
tor_tls_t.server_handshake_count we don't need that semi-broken
function (at least till there is a way to disable rfc5746
renegotiations too).
Switch 'server_handshake_count' from a uint8_t to 2 unsigned int bits.
Since we won't ever be doing more than 3 handshakes, we don't need the
extra space.
Toggle tor_tls_t.got_renegotiate based on the server_handshake_count.
Also assert that when we've done two handshakes as a server (the initial
SSL handshake, and the renegotiation handshake) we've just
renegotiated.
Finally, in tor_tls_read() return an error if we see more than 2
handshakes.
The renegotiation callback was called only when the first Application
Data arrived, instead of when the renegotiation took place.
This happened because SSL_read() returns -1 and sets the error to
SSL_ERROR_WANT_READ when a renegotiation happens instead of reading
data [0].
I also added a commented out aggressive assert that I won't enable yet
because I don't feel I understand SSL_ERROR_WANT_READ enough.
[0]: Look at documentation of SSL_read(), SSL_get_error() and
SSL_CTX_set_mode() (SSL_MODE_AUTO_RETRY section).
Introduce tor_tls_state_changed_callback(), which handles every SSL
state change.
The new function tor_tls_got_server_hello() is called every time we
send a ServerHello during a v2 handshake, and plays the role of the
previous tor_tls_server_info_callback() function.
Right now we can take the digests only of an RSA key, and only expect to
take the digests of an RSA key. The old tor_cert_get_id_digests() would
return a good set of digests for an RSA key, and an all-zero one for a
non-RSA key. This behavior is too error-prone: it carries the risk that
we will someday check two non-RSA keys for equality and conclude that
they must be equal because they both have the same (zero) "digest".
Instead, let's have tor_cert_get_id_digests() return NULL for keys we
can't handle, and make its callers explicitly test for NULL.
Our keys and x.509 certs are proliferating here. Previously we had:
An ID cert (using the main ID key), self-signed
A link cert (using a shorter-term link key), signed by the ID key
Once proposal 176 and 179 are done, we will also have:
Optionally, a presentation cert (using the link key),
signed by whomever.
An authentication cert (using a shorter-term ID key), signed by
the ID key.
These new keys are managed as part of the tls context infrastructure,
since you want to rotate them under exactly the same circumstances,
and since they need X509 certificates.
Also remove a few other related warnings that could occur during the ssl
handshake. We do this because the relay operator can't do anything about
them, and they aren't their fault.
- We were reporting the _bottom_ N failing states, not the top N.
- With bufferevents enabled, we logged all TLS states as being "in
bufferevent", which isn't actually informative.
- When we had nothing to report, we reported nothing too loudly.
- Also, we needed documentation.
This code lets us record the state of any outgoing OR connection
that fails before it becomes open, so we can notice if they're all
dying in the same SSL state or the same OR handshake state.
More work is still needed:
- We need documentation
- We need to actually call the code that reports the failure when
we realize that we're having a hard time connecting out or
making circuits.
- We need to periodically clear out all this data -- perhaps,
whenever we build a circuit successfully?
- We'll eventually want to expose it to controllers, perhaps.
Partial implementation of feature 3116.
SSL_*_app_data uses ex_data index 0, which will be the first one allocated
by SSL_get_ex_new_index. Thus, if we ever started using the ex_data feature
for some other purpose, or a library linked to Tor ever started using
OpenSSL's ex_data feature, Tor would break in spectacular and mysterious
ways. Using the SSL_*_ex_data functions directly now may save us from
that particular form of breakage in the future.
But I would not be surprised if using OpenSSL's ex_data functions at all
(directly or not) comes back to bite us on our backends quite hard. The
specified behaviour of dup_func in the man page is stupid, and
crypto/ex_data.c is a horrific mess.
Our regular DH parameters that we use for circuit and rendezvous
crypto are unchanged. This is yet another small step on the path of
protocol fingerprinting resistance.
(Backport from 0.2.2's 5ed73e3807)
Our regular DH parameters that we use for circuit and rendezvous
crypto are unchanged. This is yet another small step on the path of
protocol fingerprinting resistance.
We need filtering bufferevent_openssl so that we can wrap around
IOCP bufferevents on Windows. This patch adds a temporary option to
turn on filtering mode, so that we can test it out on non-IOCP
systems to make sure it hasn't got any surprising bugs.
It also fixes some allocation/teardown errors in using
bufferevent_openssl as a filter.
First start of a fix for bug2001, but my test network still isn't
working: the client and the server send each other VERSIONS cells,
but never notice that they got them.
* Make tor_tls_context_new internal to tortls.c, and return the new
tor_tls_context_t from it.
* Add a public tor_tls_context_init wrapper function to replace it.
This requires the latest Git version of Libevent as of 24 March 2010.
In the future, we'll just say it requires Libevent 2.0.5-alpha or
later.
Since Libevent doesn't yet support hierarchical rate limit groups,
there isn't yet support for tracking relayed-bytes separately when
using the bufferevent system. If a future version does add support
for hierarchical buckets, we can add that back in.
This should make us conflict less with system files named "log.h".
Yes, we shouldn't have been conflicting with those anyway, but some
people's compilers act very oddly.
The actual change was done with one "git mv", by editing
Makefile.am, and running
find . -name '*.[ch]' | xargs perl -i -pe 'if (/^#include.*\Wlog.h/) {s/log.h/torlog.h/; }'
Our code assumed that any version of OpenSSL before 0.9.8l could not
possibly require SSL_OP_ALLOW_UNSAFE_LEGACY_RENEGOTIATION. This is
so... except that many vendors have backported the flag from later
versions of openssl when they backported the RFC5476 renegotiation
feature.
The new behavior is particularly annoying to detect. Previously,
leaving SSL_OP_ALLOW_UNSAFE_LEGACY_RENEGOTIATION unset meant that
clients would fail to renegotiate. People noticed that one fast!
Now, OpenSSL's RFC5476 support means that clients will happily talk to
any servers there are, but servers won't accept renegotiation requests
from unpatched clients unless SSL_OP_ALLOW_etc is set. More fun:
servers send back a "no renegotiation for you!" error, which unpatched
clients respond to by stalling, and generally producing no useful
error message.
This might not be _the_ cause of bug 1346, but it is quite likely _a_
cause for bug 1346.
This time, set the SSL3_FLAGS_ALLOW_UNSAFE_RENEGOTIATION flag on every
version before OpenSSL 0.9.8l. I can confirm that the option value (0x0010)
wasn't reused until OpenSSL 1.0.0beta3.
In brief: you mustn't use the SSL3_FLAG solution with anything but 0.9.8l,
and you mustn't use the SSL_OP solution with anything before 0.9.8m, and
you get in _real_ trouble if you try to set the flag in 1.0.0beta, since
they use it for something different.
For the ugly version, see my long comment in tortls.c
We need to do this because Apple doesn't update its dev-tools headers
when it updates its libraries in a security patch. On the bright
side, this might get us out of shipping a statically linked OpenSSL on
OSX.
May fix bug 1225.
[backported]
We need to do this because Apple doesn't update its dev-tools headers
when it updates its libraries in a security patch. On the bright
side, this might get us out of shipping a statically linked OpenSSL on
OSX.
May fix bug 1225.
We were checking for msg==NULL, but not lib or proc. This case can
only occur if we have an error whose string we somehow haven't loaded,
but it's worth coding defensively here.
Spotted by rieo on IRC.
Some *_free functions threw asserts when passed NULL. Now all of them
accept NULL as input and perform no action when called that way.
This gains us consistence for our free functions, and allows some
code simplifications where an explicit null check is no longer necessary.
It turns out that OpenSSL 0.9.8m is likely to take a completely
different approach for reenabling renegotiation than OpenSSL 0.9.8l
did, so we need to work with both. :p Fixes bug 1158.
(patch by coderman; commit message by nickm)
To fix a major security problem related to incorrect use of
SSL/TLS renegotiation, OpenSSL has turned off renegotiation by
default. We are not affected by this security problem, however,
since we do renegotiation right. (Specifically, we never treat a
renegotiated credential as authenticating previous communication.)
Nevertheless, OpenSSL's new behavior requires us to explicitly
turn renegotiation back on in order to get our protocol working
again.
Amusingly, this is not so simple as "set the flag when you create
the SSL object" , since calling connect or accept seems to clear
the flags.
For belt-and-suspenders purposes, we clear the flag once the Tor
handshake is done. There's no way to exploit a second handshake
either, but we might as well not allow it.
