Previously, we used !directory_fetches_from_authorities() to predict
that we would tunnel connections. But the rules have changed
somewhat over the course of 0.2.8
So, back long ago, XXX012 meant, "before Tor 0.1.2 is released, we
had better revisit this comment and fix it!"
But we have a huge pile of such comments accumulated for a large
number of released versions! Not cool.
So, here's what I tried to do:
* 0.2.9 and 0.2.8 are retained, since those are not yet released.
* XXX+ or XXX++ or XXX++++ or whatever means, "This one looks
quite important!"
* The others, after one-by-one examination, are downgraded to
plain old XXX. Which doesn't mean they aren't a problem -- just
that they cannot possibly be a release-blocking problem.
There was a > that should have been an ==, and a missing !. These
together prevented us from issuing a warning in the case that a
nickname matched an Unnamed node only.
Fixes bug 19203; bugfix on 0.2.3.1-alpha.
There are a few places where we want to disable a warning: for
example, when it's impossible to call a legacy API without
triggering it, or when it's impossible to include an external header
without triggering it.
This pile of macros uses GCC's c99 _Pragma support, plus the usual
macro trickery, to enable and disable warnings.
Remove support for "GET /tor/bytes.txt" DirPort request, and
"GETINFO dir-usage" controller request, which were only available
via a compile-time option in Tor anyway.
Feature was added in 0.2.2.1-alpha. Resolves ticket 19035.
I introduced this bug when I moved signing_key_cert into
signed_descriptor_t. Bug not in any released Tor. Fixes bug 19175, and
another case of 19128.
Just like signed_descriptor_from_routerinfo(), routerlist_reparse_old()
copies the fields from one signed_descriptor_t to another, and then
clears the fields from the original that would have been double-freed by
freeing the original. But when I fixed the s_d_f_r() bug [#19128] in
50cbf22099, I missed the fact that the code was duplicated in
r_p_o().
Duplicated code strikes again!
For a longer-term solution here, I am not only adding the missing fix to
r_p_o(): I am also extracting the duplicated code into a new function.
Many thanks to toralf for patiently sending me stack traces until
one made sense.
If OpenSSL fails to generate an RSA key, do not retain a dangling
pointer to the previous (uninitialized) key value. The impact here
should be limited to a difficult-to-trigger crash, if OpenSSL is
running an engine that makes key generation failures possible, or if
OpenSSL runs out of memory. Fixes bug 19152; bugfix on
0.2.1.10-alpha. Found by Yuan Jochen Kang, Suman Jana, and Baishakhi
Ray.
This is potentially scary stuff, so let me walk through my analysis.
I think this is a bug, and a backport candidate, but not remotely
triggerable in any useful way.
Observation 1a:
Looking over the OpenSSL code here, the only way we can really fail in
the non-engine case is if malloc() fails. But if malloc() is failing,
then tor_malloc() calls should be tor_asserting -- the only way that an
attacker could do an exploit here would be to figure out some way to
make malloc() fail when openssl does it, but work whenever Tor does it.
(Also ordinary malloc() doesn't fail on platforms like Linux that
overcommit.)
Observation 1b:
Although engines are _allowed_ to fail in extra ways, I can't find much
evidence online that they actually _do_ fail in practice. More evidence
would be nice, though.
Observation 2:
We don't call crypto_pk_generate*() all that often, and we don't do it
in response to external inputs. The only way to get it to happen
remotely would be by causing a hidden service to build new introduction
points.
Observation 3a:
So, let's assume that both of the above observations are wrong, and the
attacker can make us generate a crypto_pk_env_t with a dangling pointer
in its 'key' field, and not immediately crash.
This dangling pointer will point to what used to be an RSA structure,
with the fields all set to NULL. Actually using this RSA structure,
before the memory is reused for anything else, will cause a crash.
In nearly every function where we call crypto_pk_generate*(), we quickly
use the RSA key pointer -- either to sign something, or to encode the
key, or to free the key. The only exception is when we generate an
intro key in rend_consider_services_intro_points(). In that case, we
don't actually use the key until the intro circuit is opened -- at which
point we encode it, and use it to sign an introduction request.
So in order to exploit this bug to do anything besides crash Tor, the
attacker needs to make sure that by the time the introduction circuit
completes, either:
* the e, d, and n BNs look valid, and at least one of the other BNs is
still NULL.
OR
* all 8 of the BNs must look valid.
To look like a valid BN, *they* all need to have their 'top' index plus
their 'd' pointer indicate an addressable region in memory.
So actually getting useful data of of this, rather than a crash, is
going to be pretty damn hard. You'd have to force an introduction point
to be created (or wait for one to be created), and force that particular
crypto_pk_generate*() to fail, and then arrange for the memory that the
RSA points to to in turn point to 3...8 valid BNs, all by the time the
introduction circuit completes.
Naturally, the signature won't check as valid [*], so the intro point
will reject the ESTABLISH_INTRO cell. So you need to _be_ the
introduction point, or you don't actually see this information.
[*] Okay, so if you could somehow make the 'rsa' pointer point to a
different valid RSA key, then you'd get a valid signature of an
ESTABLISH_INTRO cell using a key that was supposed to be used for
something else ... but nothing else looks like that, so you can't use
that signature elsewhere.
Observation 3b:
Your best bet as an attacker would be to make the dangling RSA pointer
actually contain a fake method, with a fake RSA_private_encrypt
function that actually pointed to code you wanted to execute. You'd
still need to transit 3 or 4 pointers deep though in order to make that
work.
Conclusion:
By 1, you probably can't trigger this without Tor crashing from OOM.
By 2, you probably can't trigger this reliably.
By 3, even if I'm wrong about 1 and 2, you have to jump through a pretty
big array of hoops in order to get any kind of data leak or code
execution.
So I'm calling it a bug, but not a security hole. Still worth
patching.
Fortunately, the arithmetic cannot actually overflow, so long as we
*always* check for the size of potentially hostile input before
copying it. I think we do, though. We do check each line against
MAX_LINE_LENGTH, and each object name or object against
MAX_UNPARSED_OBJECT_SIZE, both of which are 128k. So to get this
overflow, we need to have our memarea allocated way way too high up
in RAM, which most allocators won't actually do.
Bugfix on 0.2.1.1-alpha, where memarea was introduced.
Found by Guido Vranken.
Previously, if the header was present, we'd proceed even if the
function wasn't there.
Easy fix for bug 19161. A better fix would involve trying harder to
find libscrypt_scrypt.
AddressSanitizer's (ASAN) SIGSEGV handler overrides the backtrace
handler and prevents it from printing its backtrace. The output of ASAN
is different from what 'bt_test.py' expects and causes backtrace test
failures.
The 'allow_user_segv_handler' option allows applications to set their
own SIGSEGV handler but is not supported by older GCC versions. These
older GCC versions do support the 'handle_segv' which prevents ASAN from
setting its SIGSEGV handler.
Now that the field exists in signed_descriptor_t, we need to make
sure we free it when we free a signed_descriptor_t, and we need to
make sure that we don't free it when we convert a routerinfo_t to a
signed_descriptor_t.
But not in any released Tor. I found this while working on #19128.
One problem: I don't see how this could cause 19128.
We use a pretty specific pair of autoconf tests here to make sure
that we only add this code when:
a) a 64-bit signed multiply fails to link,
AND
b) the same 64-bit signed multiply DOES link correctly when
__mulodi4 is defined.
Closes ticket 19079.
We need to define this function when compiling with clang -m32 -ftrapv,
since otherwise we get link errors, since apparently some versions
of libclang_rt.builtins don't define a version of it that works? Or
clang doesn't know to look for it?
This definition is taken from the LLVM source at
https://llvm.org/svn/llvm-project/compiler-rt/trunk/lib/builtins/mulodi4.c
I've also included the license (dual BSD-ish/MIT-ish).
With the fix for #17150, I added a duplicate certificate here. Here
I remove the original location in 0.2.8. (I wouldn't want to do
that in 027, due to the amount of authority-voting-related code
drift.)
Closes 19073.
This API change makes it so that routerinfo_incompatible...() no
longer takes a routerinfo_t, so that it's obvious that it should
only look at fields from the signed_descriptor_t.
This change should prevent a recurrence of #17150.
We need this field to be in signed_descriptor_t so that
routerinfo_incompatible_with_extrainfo can work correctly (#17150).
But I don't want to move it completely in this patch, since a great
deal of the code that messes with it has been in flux since 0.2.7,
when this ticket was opened. I should open another ticket about
removing the field from routerinfo_t and extrainfo_t later on.
This patch fixes no actual behavior.
The routerinfo we pass to routerinfo_incompatible_with_extrainfo is
the latest routerinfo for the relay. The signed_descriptor_t, on
the other hand, is the signed_descriptor_t that corresponds to the
extrainfo. That means we should be checking the digest256 match
with that signed_descriptor_t, not with the routerinfo.
Fixes bug 17150 (and 19017); bugfix on 0.2.7.2-alpha.
When parsing detached signature, we make sure that we use the length of the
digest algorithm instead of an hardcoded DIGEST256_LEN in order to avoid
comparing bytes out of bound with a smaller digest length such as SHA1.
Fixes#19066
Signed-off-by: David Goulet <dgoulet@torproject.org>
We know there are overflows in curve25519-donna-c32, so we'll have
to have that one be fwrapv.
Only apply the asan, ubsan, and trapv options to the code that does
not need to run in constant time. Those options introduce branches
to the code they instrument.
(These introduced branches should never actually be taken, so it
might _still_ be constant time after all, but branch predictors are
complicated enough that I'm not really confident here. Let's aim for
safety.)
Closes 17983.
Make directory authorities write the v3-status-votes file out
to disk earlier in the consensus process, so we have the votes
even if we abort the consensus process later on.
Resolves ticket 19036.
The goal here is to provide a way to decouple pieces of the code
that want to learn "when something happens" from those that realize
that it has happened.
The implementation here consists of a generic backend, plus a set of
macros to define and implement a set of type-safe frontends.
In dirserv_compute_performance_thresholds, we allocate arrays based
on the length of 'routers', a list of routerinfo_t, but loop over
the nodelist. The 'routers' list may be shorter when relays were
filtered by routers_make_ed_keys_unique, leading to an out-of-bounds
write on directory authorities.
This bug was originally introduced in 26e89742, but it doesn't look
possible to trigger until routers_make_ed_keys_unique was introduced
in 13a31e72.
Fixes bug 19032; bugfix on tor 0.2.8.2-alpha.
This was one of our longest functions, at 600 lines. It makes a nice
table-driven URL-based function instead.
The code is a bit ugly, it leave the indentation as it is in hopes of
making pending directory.c changes easier to merge. Later we can
clean up the indentation.
Also, remove unused mallinfo export code from directory.c
Closes ticket 16698
Previously, we were using the generic schedule for some downloads,
and the consensus schedule for others.
Resolves ticket 18816; fix on fddb814fe in 0.2.4.13-alpha.
We used to be locked in to the "tap" handshake length, and now we can
handle better handshakes like "ntor".
Resolves ticket 18998.
I checked that relay_send_command_from_edge() behaves fine when you
hand it a payload with length 0. Clients behave fine too, since current
clients remain strict about the required length in the rendezvous2 cells.
(Clients will want to become less strict once they have an alternate
format that they're willing to receive.)
we should avoid launching a consensus fetch if we don't want one,
but if we do end up with an extra one, we should let the other checks
take care of it.
We'll back off from the request in connection_ap_handshake_attach_circuit,
or cancel it in connection_dir_close_consensus_fetches, and those are the
only places we need to check.
Tor stores client authorization cookies in two slightly different forms.
The service's client_keys file has the standard base64-encoded cookie,
including two chars of padding. The hostname file and the client remove
the two padding chars, and store an auth type flag in the unused bits.
The distinction makes no sense. Refactor all decoding to use the same
function, which will accept either form, and use a helper function for
encoding the truncated format.
This improves client anonymity and avoids directory header tampering.
The extra load on the authorities should be offset by the fallback
directories feature.
This also simplifies the fixes to #18809.
Delete an unnecessary check for non-preferred IP versions.
Allows clients which can't reach any directories of their
preferred IP address version to get directory documents.
Patch on #17840 in 0.2.8.1-alpha.
After #17840 in 0.2.8.1-alpha, we incorrectly chose an IPv4
address for all DIRIND_ONEHOP directory connections,
even if the routerstatus didn't have an IPv4 address.
This likely affected bridge clients with IPv6 bridges.
Resolves#18921.
The problem is that "q" is always set on the first iteration even
if the question is not a supported question. This set of "q" is
not necessary, and will be handled after exiting the loop if there
if a supported q->type was found.
[Changes file by nickm]
lease enter the commit message for your changes. Lines starting
* SHA-3/SHAKE use little endian for certain things, so byteswap as
needed.
* The code was written under the assumption that unaligned access to
quadwords is allowed, which isn't true particularly on non-Intel.
Decide to advertise begindir support in a similar way to how
we decide to advertise DirPort.
Fix up the associated descriptor-building unit tests.
Resolves#18616, bugfix on 0c8e042c30 in #12538 in 0.2.8.1-alpha.
Apparently somewhere along the line we decided that MIN might be
missing.
But we already defined it (if it was missing) in compat.h, which
everybody includes.
Closes ticket 18889.
When we connect to a hidden service as a client we may need three internal
circuits, one for the descriptor retrieval, introduction, and rendezvous.
Let's try to make sure we have them. Closes#13239.
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
Unlike tor_assert(), these macros don't abort the process. They're
good for checking conditions we want to warn about, but which don't
warrant a full crash.
This commit also changes the default implementation for
tor_fragile_assert() to tor_assert_nonfatal_unreached_once().
Closes ticket 18613.
This marks some lines as unreachable by the unit tests, and as
therefore excluded from test coverage.
(Note: This convention is only for lines that are absolutely
unreachable. Don't use it anywhere you wouldn't add a
tor_fragile_assert().)
When the directory authorities refuse a bad relay's descriptor,
encourage the relay operator to contact us. Many relay operators
won't notice this line in their logs, but it's a win if even a
few learn why we don't like what their relay was doing.
Resolves ticket 18760.
I didn't specify a contact mechanism (e.g. an email address), because
every time we've done that in the past, a few years later we noticed
that the code was pointing people to an obsolete contact address.
Also, put libor-testing.a at a better position in the list of
libraries, to avoid linker errors.
This is a fix, or part of a fix, for 18490.
Conflicts:
src/test/include.am
This changes simply renames them by removing "Testing" in front of them and
they do not require TestingTorNetwork to be enabled anymore.
Fixes#18481
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
Yes, we could cast to unsigned char first, but it's probably safest
to just use our own (in test_util), or remove bad-idea features that
we don't use (in readpassphrase.c).
Fixes 18728.
Only when we were actually flushing the cell stats to a controller
would we free them. Thus, they could stay in RAM even after the
circuit was freed (eg if we didn't have any controllers).
Fixes bug 18673; bugfix on 0.2.5.1-alpha.
Otherwise coverity complains that we're checking an whether an int64 is
less than INT64_MIN, which of course it isn't.
Fixes CID 1357176. Not in any released Tor.
They incorrectly summarized what the function was planning to do,
leading to wrong behavior like making an http request to an orport,
or making a begindir request to a dirport.
This change backs out some of the changes made in commit e72cbf7a, and
most of the changes made in commit ba6509e9.
This patch resolves bug 18625. There more changes I want to make
after this one, for code clarity.
This change allows us to simplify path selection for clients, and it
should have minimal effect in practice since >99% of Guards already have
the Stable flag. Implements ticket 18624.
Commit e72cbf7a4 introduced a change to directory_initiate_command_rend()
that made tor use the ORPort when making a directory request to the DirPort.
The primary consequence was that a relay couldn't selftest its DirPort thus
failing to work and join the network properly.
The main issue was we were always considering an anonymized connection to be
an OR connection which is not true.
Fixes#18623
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
Regardless of the setting of ExtendAllowPrivateAddresses.
This fixes a bug with pluggable transports that ignore the
(potentially private) address in their bridge line.
Fixes bug 18517; bugfix on 23b088907f in tor-0.2.8.1-alpha.
When requesting extrainfo descriptors from a trusted directory
server, check whether it is an authority or a fallback directory
which supports extrainfo descriptors.
Fixes bug 18489; bugfix on 90f6071d8d in tor-0.2.4.7-alpha.
Reported by "atagar", patch by "teor".
Make it clearer that they are about outgoing connection attempts.
Specify the options involved where they were missing from one log
message.
Clarify a comment.
Downgrade logs and backtraces about IP versions to
info-level. Only log backtraces once each time tor runs.
Assists in diagnosing bug 18351; bugfix on c3cc8e16e in
tor-0.2.8.1-alpha.
Reported by "sysrqb" and "Christian", patch by "teor".
The fd would leak when the User wasn't recogniezed by
getpwnam(). Since we'd then go on to exit, this wasn't a terribad
leak, but it's still not as nice as no leak at all.
CID 1355640; bugfix on no released Tor.
commit edeba3d4 removed a switch, but left the "break" lines in
from that switch. fortunately the resulting behavior was not wrong,
since there was an outer switch that it was ok to break from.
no actual changes here -- but the new indenting makes it clear
that the fixes in #18332 were not as good as they should have been.
the next commit will deal with that.
We've got to make sure that every single subsequent calculation in
dirserv_generate_networkstatus_vote_obj() are based on the list of
routerinfo_t *after* we've removed possible duplicates, not before.
Fortunately, none of the functions that were taking a routerlist_t
as an argument were actually using any fields other than this list
of routers.
Resolves issue 18318.DG3.
I had a half-built mechanism to track, during the voting process,
whether the Ed25519 value (or lack thereof) reflected a true
consensus among the authorities. But we never actually inserted this
field in the consensus.
The key idea here is that we first attempt to match up votes by pairs
of <Ed,RSA>, where <Ed> can be NULL if we're told that there is no
Ed key. If this succeeds, then we can treat all those votes as 'a
consensus for Ed'. And we can include all other votes with a
matching RSA key and no statement about Ed keys as being "also about
the same relay."
After that, we look for RSA keys we haven't actually found an entry
for yet, and see if there are enough votes for them, NOT considering
Ed keys. If there are, we match them as before, but we treat them
as "not a consensus about ed".
When we include an entry in a consensus, if it does not reflect a
consensus about ed keys, then we include a new NoEdConsensus flag on
it.
This is all only for consensus method 22 or later.
Also see corresponding dir-spec patch.
When generating a vote, and we have two routerinfos with the same ed
key, omit the one published earlier.
This was supposed to have been solved by key pinning, but when I
made key pinning optional, I didn't realize that this would jump up
and bite us. It is part of bug 18318, and the root cause of 17668.
If we're a server with no address configured, resolve_my_hostname
will need this. But not otherwise. And the preseeding itself can
consume a few seconds if like tails we have no resolvers.
Fixes bug 18548.
I didn't want to grant blanket permissions for chmod() and chown(),
so here's what I had to do:
* Grant open() on all parent directories of a unix socket
* Write code to allow chmod() and chown() on a given file only.
* Grant chmod() and chown() on the unix socket.
This is a part of a fix for 18253; bugfix on 0.2.8.1-alpha.
Alternatively, we could permit chmod/chown in the sandbox, but I
really don't like giving the sandbox permission to alter
permissions.
Launching 7 descriptor fetches makes a connection to each HSDir that is 6
and the seventh one fails to pick an HSDir because they are all being used
already so it was killing all pending connections at once.
Fixes#15937
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
When we made HidServDirectoryV2 always 1, we removed the situation
where a relay could choose not to be an HSDir. Now simplify the
rest of the code to reflect this decision.
(We have to remove two apparently unrelated free() calls in the unit
tests, since they used to free stuff that we created as a side effect
of calling router_get_my_routerinfo(), and now we no longer call that.)
This simplifies relay behavior, because the relay offers the hsdir
functionality independent of whether the directory authorities have
decided this relay is suitable for clients to use yet.
Implements ticket 18332.
The transproxy feature is only enabled when __FreeBSD__ is defined, and
only regular FreeBSD does that. Change this to __FreeBSD_kernel__ which
is defined on derivatives as well.
This enables the relevant options/validate__transproxy test on FreeBSD
derivatives.
Allow fallback directories which have been stable for 7 days
to work around #18050, which causes relays to submit descriptors
with 0 DirPorts when restarted. (Particularly during Tor version
upgrades.)
Ignore low fallback directory count in alpha builds.
Set the target count to 50.
On windows, you cannot open() a directory. So for Windows we should
just take our previous stat-based approach.
Closes bug 18392; bug not in any released Tor.
This is in accordance with our usual policy against freelists,
now that working allocators are everywhere.
It should also make memarea.c's coverage higher.
I also doubt that this code ever helped performance.
Short version: clang asan hates the glibc strcmp macro in
bits/string2.h if you are passing it a constant string argument of
length two or less. (I could be off by one here, but that's the
basic idea.)
Closes issue 14821.
Did you know that crypto_digest_all is a substring of
crypto_digest_alloc_bytes()? Hence the mysterious emergence of
"crypto_common_digestsoc_bytes".
Next time I should use the \b assertion in my regexen.
Spotted by Mike.
Previously, I had left in some debugging code with /*XXX*/ after it,
which nobody noticed. Live and learn! Next time I will use /*XXX
DO NOT COMMIT*/ or something.
We need to define a new consensus method for this; consensus method
21 shouldn't actually be used.
Fixes bug 17702; bugfix on 0.2.7.2-alpha.
We already write out bootstrapping progress (see bug 9927) per new
microdesc batch. There's no need to do a full "I learned some more
directory information, but not enough to..." line each time too.
Now, when a user who has set EntryNodes finishes bootstrapping, Tor
automatically repopulates the guard set based on this new directory
information. Fixes bug 16825; bugfix on 0.2.3.1-alpha.
Originally it can overflow in some weird cases. Now it should no longer
be able to do so.
Additionally, limit main's timers to 30 days rather than to 38 years;
we don't actually want any 38-year timers.
Closes bug 17682.
They are no longer "all" digests, but only the "common" digests.
Part of 17795.
This is an automated patch I made with a couple of perl one-liners:
perl -i -pe 's/crypto_digest_all/crypto_common_digests/g;' src/*/*.[ch]
perl -i -pe 's/\bdigests_t\b/common_digests_t/g;' src/*/*.[ch]
If unix socket was configured as listener (such as a ControlSocket or a
SocksPort unix socket), and tor was started as root but not configured
to switch to another user, tor would segfault while trying to string
compare a NULL value. Fixes bug 18261; bugfix on 0.2.8.1-alpha. Patch
by weasel.
Closes ticket 18242.
The rationale here is that I like having coverage on by default in my
own working directory, but I always want assertions turned on unless
I'm doing branch coverage specifically.
1. We were sometimes using libevent uninitialized, which is Not Allowed.
2. The malformed-PTR dns test was supposed to get a -1 output... but
the test was wrong, since it forgot that in-addr.arpa addresses
are in reverse order.
Bugs not in any released tor.
When ClientPreferIPv6ORPort is auto, bridges prefer the configured
bridge ORPort address. Otherwise, they use the value of the option.
Other clients prefer IPv4 ORPorts if ClientPreferIPv6ORPort is auto.
When ClientPreferIPv6DirPort is auto, all clients prefer IPv4 DirPorts.
When ClientPreferIPv6ORPort is auto, bridges prefer the configured
bridge ORPort address. Otherwise, they use the value of the option.
Other clients prefer IPv4 ORPorts if ClientPreferIPv6ORPort is auto.
When ClientPreferIPv6DirPort is auto, all clients prefer IPv4 DirPorts.
We use sensible parameters taken from common sources, and no longer
have dynamic DH groups as an option, but it feels prudent to have
OpenSSL validate p and g at initialization time.
We've never actually tested this support, and we should probably assume
it's broken.
To the best of my knowledge, only OpenVMS has this, and even on
OpenVMS it's a compile-time option to disable it. And I don't think
we build on openvms anyway. (Everybody else seems to be working
around the 2038 problem by using a 64-bit time_t, which won't expire
for roughly 292 billion years.)
Closes ticket 18184.
Modify callers to correctly handle these new NULL returns:
* fix assert in onion_extend_cpath
* warn and discard circuit in circuit_get_open_circ_or_launch
* warn, discard circuit, and tell controller in handle_control_extendcircuit
Bridge clients ignore ClientUseIPv6, acting as if it is always 1.
This preserves existing behaviour.
Make ClientPreferIPv6OR/DirPort auto by default:
* Bridge clients prefer IPv6 by default.
* Other clients prefer IPv4 by default.
This preserves existing behaviour.
ClientUseIPv4 0 tells tor to avoid IPv4 client connections.
ClientPreferIPv6DirPort 1 tells tor to prefer IPv6 directory connections.
Refactor policy for IPv4/IPv6 preferences.
Fix a bug where node->ipv6_preferred could become stale if
ClientPreferIPv6ORPort was changed after the consensus was loaded.
Update documentation, existing code, add unit tests.
node_get_all_orports and router_get_all_orports incorrectly used or_port
with IPv6 addresses. They now use ipv6_orport.
Also refactor and remove duplicated code.
This closes bug 18162; bugfix on a45b131590, which fixed a related
issue long ago.
In addition to the #18162 issues, this fixes a signed integer overflow
in smarltist_add_all(), which is probably not so great either.
Avoid using a pronoun where it makes comments unclear.
Avoid using gender for things that don't have it.
Avoid assigning gender to people unnecessarily.
Allow fallback directories which have been stable for 30 days
to work around #18050, which causes relays to submit descriptors
with 0 DirPorts when restarted. (Particularly during Tor version
upgrades.)
Ignore low fallback directory count in alpha builds.
Check size argument to memwipe() for underflow.
Closes bug #18089. Reported by "gk", patch by "teor".
Bugfix on 0.2.3.25 and 0.2.4.6-alpha (#7352),
commit 49dd5ef3 on 7 Nov 2012.
Otherwise, relays publish a descriptor with DirPort 0 when the DirPort
reachability test takes longer than the ORPort reachability test.
Closes bug #18050. Reported by "starlight", patch by "teor".
Bugfix on 0.1.0.1-rc, commit a1f1fa6ab on 27 Feb 2005.
Sometimes you can call time() and then touch a file, and have the
second come out a little before the first. See #18025 for way more
information than you necessarily wanted.
LibreSSL doesn't use OpenSSL_version (it uses the older SSLeay_version
API), but it reports a major version number as 2 in
OPENSSL_VERSION_NUMBER. Instead of fudging the version check, for now,
let's just check if we're using LibreSSL by checking the version number
macro exists, and use compatibility defines unconditionally when we
detect LibreSSL.
When _list() is called with AF_UNSPEC family and fails to enumerate
network interfaces using platform specific API, have it call
_hack() twice to find out IPv4 and/or IPv6 address of a machine Tor
instance is running on. This is correct way to handle this case
because _hack() can only be called with AF_INET and AF_INET6 and
does not support any other address family.
OpenSSL doesn't use them, and fwict they were never called. If some
version of openssl *does* start using them, we should test them before
we turn them back on.
See ticket 17926
This creates a random 100 KiB buffer, and incrementally hashes
(SHA3-512) between 1 and 5 * Rate bytes in a loop, comparing the running
digest with the equivalent one shot call from the start of the buffer.
This is an eXtendable-Output Function with the following claimed
security strengths against *all* adversaries:
Collision: min(d/2, 256)
Preimage: >= min(d, 256)
2nd Preimage: min(d, 256)
where d is the amount of output used, in bits.
* DIGEST_SHA3_[256,512] added as supported algorithms, which do
exactly what is said on the tin.
* test/bench now benchmarks all of the supported digest algorithms,
so it's possible to see just how slow SHA-3 is, though the message
sizes could probably use tweaking since this is very dependent on
the message size vs the SHA-3 rate.
The digest routines use init/update/sum, where sum will automatically
copy the internal state to support calculating running digests.
The XOF routines use init/absorb/squeeze, which behave exactly as stated
on the tin.
This will give relay operators the ability of disabling the caching of
directory data. In general, this should not be necessary, but on some
lower-resource systems it may beneficial.
I believe that the final SMARTLIST_DEL_CURRENT was sometimes
double-removing items that had already been removed by
connection_mark_unattached_ap or
connection_ap_handshake_attach_circuit().
The fix here is to prevent iteration over the list that other
functions might be modifying.
According to the POSIX standard the option value is a pointer to void
and the option length a socklen_t. The Windows implementation makes the
option value be a pointer to character and the option length an int.
Casting the option value to a pointer to void conforms to the POSIX
standard while the implicit cast to a pointer to character conforms to
the Windows implementation.
The casts of the option length to the socklen_t data type conforms to
the POSIX standard. The socklen_t data type is actually an alias of an
int so it also conforms to the Windows implementation.
It is AP-specific, so that's where it belongs. This shouldn't have
caused a bug, but due to #17876, we were never actually calling
connection_edge_about_to_close from connection_ap_about_to_close,
causing bug #17874 (aka bug #17752).
When a relay does not have an open directory port but it has an
orport configured and is accepting client connections then it can
now service tunnelled directory requests, too. This was already true
of relays with an dirport configured.
We also conditionally stop advertising this functionality if the
relay is nearing its bandwidth usage limit - same as how dirport
advertisement is determined.
Partial implementation of prop 237, ticket 12538
Applies the 6c443e987d fix to router_pick_directory_server_impl.
6c443e987d applied to directory servers chosen from the consensus,
and was:
"Tweak the 9969 fix a little
If we have busy nodes and excluded nodes, then don't retry with the
excluded ones enabled. Instead, wait for the busy ones to be nonbusy."
These IPv6 addresses must be quoted, because : is the port separator,
and "acce" is a valid hex block.
Add unit tests for assumed actions in IPv6 policies.
"Tor has included a feature to fetch the initial consensus from nodes
other than the authorities for a while now. We just haven't shipped a
list of alternate locations for clients to go to yet.
Reasons why we might want to ship tor with a list of additional places
where clients can find the consensus is that it makes authority
reachability and BW less important.
We want them to have been around and using their current key, address,
and port for a while now (120 days), and have been running, a guard,
and a v2 directory mirror for most of that time."
Features:
* whitelist and blacklist for an opt-in/opt-out trial.
* excludes BadExits, tor versions that aren't recommended, and low
consensus weight directory mirrors.
* reduces the weighting of Exits to avoid overloading them.
* places limits on the weight of any one fallback.
* includes an IPv6 address and orport for each FallbackDir, as
implemented in #17327. (Tor won't bootstrap using IPv6 fallbacks
until #17840 is merged.)
* generated output includes timestamps & Onionoo URL for traceability.
* unit test ensures that we successfully load all included default
fallback directories.
Closes ticket #15775. Patch by "teor".
OnionOO script by "weasel", "teor", "gsathya", and "karsten".
* The option is now KeepBindCapabilities
* We now warn if the user specifically asked for KeepBindCapabilities
and we can't deliver.
* The unit tests are willing to start.
* Fewer unused-variable warnings.
* More documentation, fewer misspellings.
router_digest_is_fallback_dir returns 1 if the digest is in the
currently loaded list of fallback directories, and 0 otherwise.
This function is for future use.
Once tor is downloading a usable consensus, any other connection
attempts are not needed.
Choose a connection to keep, favouring:
* fallback directories over authorities,
* connections initiated earlier over later connections
Close all other connections downloading a consensus.
Prop210: Add attempt-based connection schedules
Existing tor schedules increment the schedule position on failure,
then retry the connection after the scheduled time.
To make multiple simultaneous connections, we need to increment the
schedule position when making each attempt, then retry a (potentially
simultaneous) connection after the scheduled time.
(Also change find_dl_schedule_and_len to find_dl_schedule, as it no
longer takes or returns len.)
Prop210: Add multiple simultaneous consensus downloads for clients
Make connections on TestingClientBootstrapConsensus*DownloadSchedule,
incrementing the schedule each time the client attempts to connect.
Check if the number of downloads is less than
TestingClientBootstrapConsensusMaxInProgressTries before trying any
more connections.
UseDefaultFallbackDirs enables any hard-coded fallback
directory mirrors. Default is 1, set it to 0 to disable fallbacks.
Implements ticket 17576.
Patch by "teor".
On FreeBSD backtrace(3) uses size_t instead of int (as glibc does). This
causes integer precision loss errors when we used int to store its
results.
The issue is fixed by using size_t to store the results of backtrace(3).
The manual page of glibc does not mention that backtrace(3) returns
negative values. Therefore, no unsigned integer wrapping occurs when its
result is stored in an unsigned data type.
Using variables removes the ambiguity about when to use variables and
when to use substitutions. Variables always work. Substitutions only
work when Autoconf knows about them which is not always the case.
The variables are also placed between quotes to ensures spaces in the
variables are handled properly.
Update the code for IPv6 authorities and fallbacks for function
argument changes.
Update unit tests affected by the function argument changes in
the patch.
Add unit tests for authority and fallback:
* adding via a function
* line parsing
* adding default authorities
(Adding default fallbacks is unit tested in #15775.)
The internal memory allocation and history object counters of the
reputation code can be used to verify the correctness of (part of) the
code. Using these counters revealed an issue where the memory allocation
counter is not decreased when the bandwidth arrays are freed.
A new function ensures the memory allocation counter is decreased when a
bandwidth array is freed.
This commit also removes an unnecessary cast which was found while
working on the code.
* Since the variable is no longer modified, it should be called
'policy' instead of 'dest'. ("Dest" is short for
"destination".)
* Fixed the space issue that dgoulet found on the ticket.
* Fixed the comment a little. (We use the imperative for function
documentation.)
Some functions that use digest maps did not mention that the digests are
expected to have DIGEST_LEN bytes. This lead to buffer over-reads in the
past.
The hidden service descriptor cache (rendcache) tests use digest maps
which expect keys to have a length of DIGEST_LEN.
Because the tests use key strings with a length lower than DIGEST_LEN,
the internal copy operation reads outside the key strings which leads to
buffer over-reads.
The issue is resolved by using character arrays with a size of
DIGEST_LEN.
Patch on ade5005853.
The tests pass empty digest strings to the dir_server_new function which
copies it into a directory server structure. The copy operation expects
the digest strings to be DIGEST_LEN characters long.
Because the length of the empty digest strings are lower than
DIGEST_LEN, the copy operation reads outside the digest strings which
leads to buffer over-reads.
The issue is resolved by using character arrays with a size of
DIGEST_LEN.
Patch on 4ff08bb581.
When an HS process an INTRODUCE2 cell, we didn't validate if the IP address
of the rendezvous point was a local address. If it's the case, we end up
wasting resources by trying to extend to a local address which fails since
we do not allow that in circuit_extend().
This commit now rejects a rendezvous point that has a local address once
seen at the hidden service side unless ExtendAllowPrivateAddresses is set.
Fixes#8976
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
There was a dead check when we made sure that an array member of a
struct was non-NULL. Tor has been doing this check since at least
0.2.3, maybe earlier.
Fixes bug 17781.
Previously we'd suppressed the mask-bits field in the output when
formatting a policy if it was >=32. But that should be a >=128 if
we're talking about IPv6.
Since we didn't put these in descriptors, this bug affects only log
messages and controller outputs.
Fix for bug 16056. The code in question was new in 0.2.0, but the
bug was introduced in 0.2.4 when we started supporting IPv6 exits.
port is in host order (addr is tor_addr_t, endianness is abstracted).
addr and port can be different to conn->addr and conn->port if
connecting via a proxy.
Consistently ignore multicast addresses when automatically
generating reject private exit policies.
Closes ticket 17763. Bug fix on 10a6390deb,
not in any released version of Tor. Patch by "teor".
The tor_cert_get_checkable_sig function uses the signing key included in
the certificate (if available) when a separate public key is not given.
When the signature is valid, the tor_cert_checksig function copies the
public key from the checkable structure to the public key field of the
certificate signing key.
In situations where the separate public key is not given but the
certificate includes a signing key, the source and destination pointers
in the copy operation are equal and invoke undefined behavior.
Undefined behaviour is avoided by ensuring both pointers are different.
These functions must really never fail; so have crypto_rand() assert
that it's working okay, and have crypto_seed_rng() demand that
callers check its return value. Also have crypto_seed_rng() check
RAND_status() before returning.
Stop ignoring ExitPolicyRejectPrivate in getinfo
exit-policy/reject-private. Fix a memory leak.
Set ExitPolicyRejectPrivate in the unit tests, and make a mock
function declaration static.
Fix unit tests for get_interface_address6_list to assume less
about the interface addresses on the system.
Instead, mock get_interface_address6_list and use the mocked
function to provide a range of address combinations.
(If we take the branch above this assertion, than we *didn't* have a
v1 handshake. So if we don't take the branch, we did. So if we
reach this assertion, we must be running as a server, since clients
no longer attempt v1 handshakes.)
Fix for bug 17654; bugfix on 9d019a7db7.
Bug not in any released Tor.
Refuse connection requests to private OR addresses unless
ExtendAllowPrivateAddresses is set. Previously, tor would
connect, then refuse to send any cells to a private address.
Fixes bugs 17674 and 8976; bugfix on b7c172c9ec (28 Aug 2012)
Original bug 6710, released in 0.2.3.21-rc and an 0.2.2 maint
release.
Patch by "teor".
This migrates away from SHA1, and provides further hash flooding
protection on top of the randomised siphash implementation.
Add unit tests to make sure that different inputs don't have the
same hash.
The wrong list was used when looking up expired intro points in a rend
service object causing what we think could be reachability issues and
triggering a BUG log.
Fixes#16702
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
exit-policy/reject-private lists the reject rules added by
ExitPolicyRejectPrivate. This makes it easier for stem to
display exit policies.
Add unit tests for getinfo exit-policy/*.
Completes ticket #17183. Patch by "teor".
Modify policies_parse_exit_policy_reject_private so it also blocks
the addresses configured for OutboundBindAddressIPv4_ and
OutboundBindAddressIPv6_, and any publicly routable port addresses
on exit relays.
Add and update unit tests for these functions.
Move the code that rejects publicly routable exit relay addresses
to policies_parse_exit_policy_reject_private. Add
addr_policy_append_reject_addr_list and use it to reject interface
addresses.
This removes the duplicate reject checks on local_address and
ipv6_local_address, but duplicates will be removed by
exit_policy_remove_redundancies at the end of the function.
This also removes the info-level logging on rejected interface
addresses. Instead, log a debug-level message in
addr_policy_append_reject_addr.
This simplifies policies_parse_exit_policy_internal and prepares for
reporting these addresses over the control port in #17183.
In my testing, an IPv6-only FreeBSD jail without ::1 returned EINVAL
from tor_ersatz_socketpair. Let's not fail the unit test because of
this - it would only ever use tor_socketpair() anyway.
(But it won't work on some systems without IPv4/IPv6 localhost
(some BSD jails) by design, to avoid creating sockets on routable
IP addresses. However, those systems likely have the AF_UNIX socketpair,
which tor prefers.)
Fixes bug #17638; bugfix on a very early tor version,
earlier than 22dba27d8d (23 Nov 2004) / svn:r2943.
Patch by "teor".
Make unit tests pass on IPv6-only systems, and systems without
localhost addresses (like some FreeBSD jails).
Fixes:
* get_if_addrs_ifaddrs: systems without localhost
* get_if_addrs_ioctl: only works on IPv4 systems
* socket: check IPv4 and IPv6, skip on EPROTONOSUPPORT
* socketpair_ersatz: uses IPv4, skip on EPROTONOSUPPORT
Fixes bug #17632; bugfix on unit tests in 0.2.7.3-rc.
c464a36772 was a partial fix for this issue in #17255;
it was released in unit tests in 0.2.7.4-rc.
Patch by "teor".
Make unit tests pass on IPv6-only systems, and systems without
localhost addresses (like some FreeBSD jails).
Fixes:
* get_if_addrs_ifaddrs: systems without localhost
* get_if_addrs_ioctl: only works on IPv4 systems
* socket: check IPv4 and IPv6, skip on EPROTONOSUPPORT
* socketpair_ersatz: uses IPv4, skip on EPROTONOSUPPORT
Fixes bug #17632; bugfix on unit tests in 0.2.7.3-rc.
c464a36772 was a partial fix for this issue in #17255;
it was released in unit tests in 0.2.7.4-rc.
Patch by "teor".
Loading a on disk bridge descriptor causes a directory download to be
scheduled, which asserts due to the periodic events not being
initialized yet.
Fixes bug #17635, not in any released version of tor.
Now we only re-scan the list in the cases we did before: when we
have a new circuit that we should try attaching to, or when we have
added a new stream that we haven't tried to attach yet.
This is part of 17590.
Long ago we used to call connection_ap_handshake_attach_circuit()
only in a few places, since connection_ap_attach_pending() attaches
all the pending connections, and does so regularly. But this turned
out to have a performance problem: it would introduce a delay to
launching or connecting a stream.
We couldn't just call connection_ap_attach_pending() every time we
make a new connection, since it walks the whole connection list. So
we started calling connection_ap_attach_pending all over, instead!
But that's kind of ugly and messes up our callgraph.
So instead, we now have connection_ap_attach_pending() use a list
only of the pending connections, so we can call it much more
frequently. We have a separate function to scan the whole
connection array to see if we missed adding anything, and log a
warning if so.
Closes ticket #17590
Mark fallback directory mirrors as "too busy" when they return
a 503 response. Previously, the code just marked authorities as busy.
Unless clients set their own fallback directories, they will never see
this bug. (There are no default fallbacks yet.)
Fixes bug 17572; bugfix on 5c51b3f1f0 released in 0.2.4.7-alpha.
Patch by "teor".
