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.
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.
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.
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]
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.
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.
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.
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.
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 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 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.
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.
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.
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".
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.
(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".
Without this check, we potentially look up to 3 characters before
the start of a malloc'd segment, which could provoke a crash under
certain (weird afaik) circumstances.
Fixes 17404; bugfix on 0.2.6.3-alpha.
Since 11150 removed client-side support for renegotiation, we no
longer need to make sure we have an openssl/TLSvX combination that
supports it (client-side)
Now that x509_get_not{Before,After} are functions in OpenSSL 1.1
(not yet releasesd), we need to define a variant that takes a const
pointer to X509 and returns a const pointer to ASN1_time.
Part of 17237. I'm not convinced this is an openssl bug or a tor
bug. It might be just one of those things.
When logging to syslog, allow a tag to be added to the syslog identity
("Tor"), i.e. the string prepended to every log message. The tag can be
configured by setting SyslogIdentityTag and defaults to none. Setting
it to "foo" will cause logs to be tagged as "Tor-foo". Closes: #17194.
Ensure that either a valid address is returned in address pointers,
or that the address data is zeroed on error.
Ensure that free_interface_address6_list handles NULL lists.
Add unit tests for get_interface_address* failure cases.
Fixes bug #17173.
Patch by fk/teor, not in any released version of tor.
... that was removed by 31eb486c46 which first appeared in
0.2.7.3-rc.
If tor is running in a ElectroBSD (or FreeBSD) jail it can't
get any IP addresses that aren't assigned to the jail by
looking at the interfaces and (by design) the
get_interface_address6_via_udp_socket_hack() fallback doesn't
work either.
The missing return code check resulted in tor_addr_is_internal()
complaining about a "non-IP address of type 49", due to reading
uninitialised memory.
Fixes#17173.
src/test/test_policy.c:
Merged calls to policies_parse_exit_policy by adding additional arguments.
fixup to remaining instance of ~EXIT_POLICY_IPV6_ENABLED.
Compacting logic test now produces previous list length of 4, corrected this.
src/config/torrc.sample.in:
src/config/torrc.minimal.in-staging:
Merged torrc modification dates in favour of latest.
ExitPolicyRejectPrivate now rejects more local addresses by default:
* the relay's published IPv6 address (if any), and
* any publicly routable IPv4 or IPv6 addresses on any local interfaces.
This resolves a security issue for IPv6 Exits and multihomed Exits that
trust connections originating from localhost.
Resolves ticket 17027. Patch by "teor".
Patch on 42b8fb5a15 (11 Nov 2007), released in 0.2.0.11-alpha.
In previous versions of Tor, ExitPolicy accept6/reject6 * produced
policy entries for IPv4 and IPv6 wildcard addresses.
To reduce operator confusion, change accept6/reject6 * to only produce
an IPv6 wildcard address.
Resolves bug #16069.
Patch on 2eb7eafc9d and a96c0affcb (25 Oct 2012),
released in 0.2.4.7-alpha.
When parsing torrc ExitPolicies, we now warn if:
* an IPv4 address is used on an accept6 or reject6 line. The line is
ignored, but the rest of the policy items in the list are used.
(accept/reject continue to allow both IPv4 and IPv6 addresses in torrcs.)
* a "private" address alias is used on an accept6 or reject6 line.
The line filters both IPv4 and IPv6 private addresses, disregarding
the 6 in accept6/reject6.
When parsing torrc ExitPolicies, we now issue an info-level message:
* when expanding an accept/reject * line to include both IPv4 and IPv6
wildcard addresses.
In each instance, usage advice is provided to avoid the message.
Partial fix for ticket 16069. Patch by "teor".
Patch on 2eb7eafc9d and a96c0affcb (25 Oct 2012),
released in 0.2.4.7-alpha.
Add get_interface_address[6]_list by refactoring
get_interface_address6. Add unit tests for new and existing functions.
Preparation for ticket 17027. Patch by "teor".
Patch on 42b8fb5a15 (11 Nov 2007), released in 0.2.0.11-alpha.
(These inputs are possible when Shadow starts the world at time_t 0,
and breaks our assumption that Tor didn't exist in the 1970s.)
Fixes regression introduced in 241e6b09. Fixes#16980.
The code was always in our Ed25519 wrappers, so enable it when using
the ed25519-donna backend, and deal with the mocking related
crypto_rand silliness.
Implements feature 16533.
If setrlimit() failed, max_out wasn't set in set_max_file_descriptors()
ending in a state where we don't use ULIMIT_BUFFER for things like tor
private key files.
Also fix the set_max_file_descriptors() documentation.
Fixes#16274
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
According to POSIX, the mutex must be locked by the thread calling the signal
functions to ensure predictable scheduling behavior.
Found the issue using Helgrind which gave the warning `dubious: associated lock
is not held by any thread`.
The base64 and base32 functions used to be in crypto.c;
crypto_format.h had no header; some general-purpose functions were in
crypto_curve25519.c.
This patch makes a {crypto,util}_format.[ch], and puts more functions
there. Small modules are beautiful!
The control port was using set_max_file_descriptors() with a limit set to 0
to query the number of maximum socket Tor can use. With the recent changes
to that function, a check was introduced to make sure a user can not set a
value below the amount we reserved for non socket.
This commit adds get_max_sockets() that returns the value of max_sockets so
we can stop using that "setter" function to get the current value.
Finally, the dead code is removed that is the code that checked for limit
equal to 0. From now on, set_max_file_descriptors() should never be used
with a limit set to 0 for a valid use case.
Fixes#16697
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
URI syntax (and DNS syntax) allows for a single trailing `.` to
explicitly distinguish between a relative and absolute
(fully-qualified) domain name. While this is redundant in that RFC 1928
DOMAINNAME addresses are *always* fully-qualified, certain clients
blindly pass the trailing `.` along in the request.
Fixes bug 16674; bugfix on 0.2.6.2-alpha.
1) We already require C99.
2) This allows us to support MSVC again (thanks to Gisle Vanem for
this part)
3) This change allows us to dump some rotten old compatibility code
from log.c
The runtime sanity checking is slightly different from the optimized
basepoint stuff in that it uses a given implementation's self tests if
available, and checks if signing/verification works with a test vector
from the IETF EdDSA draft.
The unit tests include a new testcase that will fuzz donna against ref0,
including the blinding and curve25519 key conversion routines. If this
is something that should be done at runtime (No?), the code can be
stolen from there.
Note: Integrating batch verification is not done yet.
Integration work scavanged from nickm's `ticket8897_9663_v2` branch,
with minor modifications. Tor will still sanity check the output but
now also attempts to catch extreme breakage by spot checking the
optimized implementation vs known values from the NaCl documentation.
Implements feature 9663.
Integrate ed25519-donna into the build process, and provide an
interface that matches the `ref10` code. Apart from the blinding and
Curve25519 key conversion, this functions as a drop-in replacement for
ref10 (verified by modifying crypto_ed25519.c).
Tests pass, and the benchmarks claim it is quite a bit faster, however
actually using the code requires additional integration work.
RFC 952 is approximately 30 years old, and people are failing to comply,
by serving A records with '_' as part of the hostname. Since relaxing
the check is a QOL improvement for our userbase, relax the check to
allow such abominations as destinations, especially since there are
likely to be other similarly misconfigured domains out there.
clang complains that the address of struct member in an assert in
SSL_SESSION_get_master_key is always non-NULL.
Instead, check each pointer argument is non-NULL before using it.
Fix on f90a704f12 from 27 May 2015, not in any released version of tor.
Unused variable warnings were still generated under some versions of OpenSSL.
Instead, make sure all variables are used under all versions.
Fix on 496df21c89, not in any released version of tor.
An earlier version of these tests was broken; now they're a nicer,
more robust, more black-box set of tests. The key is to have each
test check a handshake message that is wrong in _one_ way.
When there are annotations on a router descriptor, the
ed25519-identity element won't be at position 0 or 1; it will be at
router+1 or router-1.
This patch also adds a missing smartlist function to search a list for
an item with a particular pointer.
Routers now use TAP and ntor onion keys to sign their identity keys,
and put these signatures in their descriptors. That allows other
parties to be confident that the onion keys are indeed controlled by
the router that generated the descriptor.
For prop220, we have a new ed25519 certificate type. This patch
implements the code to create, parse, and validate those, along with
code for routers to maintain their own sets of certificates and
keys. (Some parts of master identity key encryption are done, but
the implementation of that isn't finished)
If the OpenSSL team accepts my patch to add an
SSL_get_client_ciphers function, this patch will make Tor use it
when available, thereby working better with openssl 1.1.
We previously used this function instead of SSL_set_cipher_list() to
set up a stack of client SSL_CIPHERs for these reasons:
A) In order to force a particular order of the results.
B) In order to be able to include dummy entries for ciphers that
this build of openssl did not support, so we could impersonate
Firefox harder.
But we no longer do B, since we merged proposal 198 and stopped
lying about what ciphers we know.
And A was actually pointless, since I had misread the implementation
of SSL_set_cipher_list(). It _does_ do some internal sorting, but
that is pre-sorting on the master list of ciphers, not sorting on
the user's preferred order.
As OpenSSL >= 1.0.0 is now required, ECDHE is now mandatory. The group
has to be validated at runtime, because of RedHat lawyers (P224 support
is entirely missing in the OpenSSL RPM, but P256 is present and is the
default).
Resolves ticket #16140.
The key here is to never touch ssl->cipher_list directly, but only
via SSL_get_ciphers(). But it's not so simple.
See, if there is no specialized cipher_list on the SSL object,
SSL_get_ciphers returns the cipher_list on the SSL_CTX. But we sure
don't want to modify that one! So we need to use
SSL_set_cipher_list first to make sure that we really have a cipher
list on the SSL object.
This field was only needed to work with the now-long-gone (I hope,
except for some horrible apples) openssl 0.9.8l; if your headers say
you have openssl 1.1, you won't even need it.
OpenSSL 1.1.0 must be built with "enable-deprecated", and compiled with
`OPENSSL_USE_DEPRECATED` for this to work, so instead, use the newer
routine as appropriate.
Use it in the sample_laplace_distribution function to make sure we return
the correct converted value after math operations are done on the input
values.
Thanks to Yawning for proposing a solution.
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
- Rewrite changes file.
- Avoid float comparison with == and use <= instead.
- Add teor's tor_llround(trunc(...)) back to silence clang warnings.
- Replace tt_assert() with tt_i64_op() and friends.
- Fix whitespace and a comment.
Consistently check for overflow in round_*_to_next_multiple_of.
Check all round_*_to_next_multiple_of functions with expected values.
Check all round_*_to_next_multiple_of functions with maximal values.
Related to HS stats in #13192.
Avoid division by zero.
Avoid taking the log of zero.
Silence clang type conversion warnings using round and trunc.
The existing values returned by the laplace functions do not change.
Add tests for laplace edge cases.
These changes pass the existing unit tests without modification.
Related to HS stats in #13192.
These commands allow for the creation and management of ephemeral
Onion ("Hidden") services that are either bound to the lifetime of
the originating control connection, or optionally the lifetime of
the tor instance.
Implements #6411.
Incidently, this fixes a bug where the maximum value was never used when
only using crypto_rand_int(). For instance this example below in
rendservice.c never gets to INTRO_POINT_LIFETIME_MAX_SECONDS.
int intro_point_lifetime_seconds =
INTRO_POINT_LIFETIME_MIN_SECONDS +
crypto_rand_int(INTRO_POINT_LIFETIME_MAX_SECONDS -
INTRO_POINT_LIFETIME_MIN_SECONDS);
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
This uses a Linux-ism to attempt to always clean up background processes
if possible. Note that it is not a catch-all, in that executables with
suid/sgid or elevated capabilities will have the prctl() attribute
stripped as part of the execve().
Resolves ticket 15471.
Background processes spawned by Tor now will have a valid stdin.
Pluggable transports can detect this behavior with the aformentioned
enviornment variable, and exit if stdin ever gets closed.
On clang (and elsewhere?) __PRETTY_FUNCTION__ includes parenthesized
argument lists. This is clever, but it makes our old "%s(): " format
look funny.
This is a fix on 0957ffeb, aka svn:r288. Fixes bug 15269.
Remove src/or/or_sha1.i and src/common/common_sha1.i on `make clean` and remove
the temporary micro-revision file when its no longer needed.
Additional changes;
- show a message when generating the micro-revision file.
- add the temporary micro revision file to the list of files to be removed on
`make clean` just in case.
- fix indentation of the make rule to improve readability.
When calling pthread_attr_setdetachstate, we were using 1 as the
argument. But the pthreads documentation says that you have to say
PTHREAD_CREATE_DETACH, which on Solaris is apparently 0x40. Calling
pthread_attr_setdetachstate with 1 crashes on Solaris with FLTBOUNDS.
(Because we're so late in the release cycle, I made the code define
PTHREAD_CREATE_DETACHED if it doesn't exist, so we aren't likely to
break any other platforms.)
This bug was introduced when we made threading mandatory in
0.2.6.1-alpha; previously, we had force-disabled threading on
Solaris. See #9495 discussion.
__libc_message() tries to open /dev/tty with O_RDWR, but the sandbox
catches that and calls it a crash. Instead, I'm making the sandbox
setenv LIBC_FATAL_STDERR_, so that glibc uses stderr instead.
Fix for 14759, bugfix on 0.2.5.1-alpha
They have been off-by-default since 0.2.5 and nobody has complained. :)
Also remove the buf_shrink() function, which hasn't done anything
since we first stopped using contiguous memory to store buffers.
Closes ticket 14848.
This both fixes the problem, and ensures that forgetting to update
domain_list in the future will trigger the bug codepath instead of
a NULL pointer deref.
If the returned value of read/recv is 0 (meaning EOF), we'll end up in an
infinite loop (active wait) until something is written on the pipe which is
not really what we want here especially because those functions are called
from the main thread.
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
Introduces two new circuit status name-value parameters: SOCKS_USERNAME
and SOCKS_PASSWORD. Values are enclosing in quotes and unusual characters
are escaped.
Example:
650 CIRC 5 EXTENDED [...] SOCKS_USERNAME="my_username" SOCKS_PASSWORD="my_password"
This is to avoid that the pthread_cond_timedwait() is not affected by time
adjustment which could make the waiting period very long or very short which
is not what we want in any cases.
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
Previously I used one queue per worker; now I use one queue for
everyone. The "broadcast" code is gone, replaced with an idempotent
'update' operation.
This way we can use the linux eventfd extension where available.
Using EVFILT_USER on the BSDs will be a teeny bit trickier, and will
require libevent hacking.
Also, re-enable the #if'd out condition-variable code.
Work queues are going to make us hack on all of this stuff a bit more
closely, so it might not be a terrible idea to make it easier to hack.
When tor is configured with --enable-bufferevents, the build fails
because compat_libevent.h makes use of the macro MOCK_DECL() which
is defined in testsupport.h, but not included. We add the include.
apparantly, "pragma GCC diagnostic push/pop" don't exist with older versions.
Fixes bug in 740e592790f570c446cbb5e6d4a77f842f75; bug not in any
released Tor.
Fixes bug 11454, where we would keep around a superseded descriptor
if the descriptor replacing it wasn't at least a week later. Bugfix
on 0.2.1.8-alpha.
Fixes bug 11457, where a certificate with a publication time in the
future could make us discard existing (and subsequent!) certificates
with correct publication times. Bugfix on 0.2.0.3-alpha.
Have clients and authorities both have new behavior, since the
fix for bug 11243 has gone in. But make clients still accept
accept old bogus HSDir descriptors, to avoid fingerprinting trickery.
Fixes bug 9286.
Otherwise we risk a subsequent memdup or memcpy copying
uninitialized RAM into some other place that might eventually expose
it. Let's make sure that doesn't happen.
Closes ticket 14041
When receiving a trasnsparently proxied request with tor using iptables tor
dies because the appropriate getsockopt calls aren't enabled on the sandbox.
This patch fixes this by adding the two getsockopt calls used when doing
transparent proxying with tor to the sandbox for the getsockopt policy.
This patch is released under the same license as the original file as
long as the author is credited.
Signed-off-by: Francisco Blas Izquierdo Riera (klondike) <klondike@gentoo.org>
This is a good idea in case the caller stupidly doesn't check the
return value from baseX_decode(), and as a workaround for the
current inconsistent API of base16_decode.
Prevents any fallout from bug 14013.
The address of an array in the middle of a structure will
always be non-NULL. clang recognises this and complains.
Disable the tautologous and redundant check to silence
this warning.
A comment about an IPv6 address string incorrectly refers
to an IPv4 address format.
A log buffer is sized 10024 rather than 10240.
Fixes bug 14001.
We add a compression level argument to tor_zlib_new, and use it to
determine how much memory to allocate for the zlib object. We use the
existing level by default, but shift to smaller levels for small
requests when we have been over 3/4 of our memory usage in the past
half-hour.
Closes ticket 11791.
Instead, generate new keys, and overwrite the empty key files.
Adds FN_EMPTY to file_status_t and file_status.
Fixes bug 13111.
Related changes due to review of FN_FILE usage:
Stop generating a fresh .old RSA key file when the .old file is missing.
Avoid overwriting .old key files with empty key files.
Skip loading zero-length extra info store, router store, stats, state,
and key files.
Stop crashing when a NULL filename is passed to file_status(),
instead, return FN_ERROR.
Also return FN_ERROR when a zero-length filename is passed to file_status().
Fixed as part of bug 13111.
Silence clang warnings under --enable-expensive-hardening, including:
+ implicit truncation of 64 bit values to 32 bit;
+ const char assignment to self;
+ tautological compare; and
+ additional parentheses around equality tests. (gcc uses these to
silence assignment, so clang warns when they're present in an
equality test. But we need to use extra parentheses in macros to
isolate them from other code).
By now, support in the network is widespread and it's time to require
more modern crypto on all Tor instances, whether they're clients or
servers. By doing this early in 0.2.6, we can be sure that at some point
all clients will have reasonable support.
Ensure we securely wipe keys from memory after
crypto_digest_get_digest and init_curve25519_keypair_from_file
have finished using them.
Fixes bug 13477.
Check all date/time values passed to tor_timegm
and parse_rfc1123_time for validity, taking leap
years into account.
Improves HTTP header validation.
Avoid unlikely signed integer overflow in tor_timegm
on systems with 32-bit time_t.
Fixes bug 13476.
Fix an instance of integer overflow in format_time_interval() when
taking the absolute value of the supplied signed interval value.
Fixes bug 13393.
Create unit tests for format_time_interval().
Fixes bug 13295; bugfix on 0.2.5.3-alpha.
The alternative here is to call crypto_global_init() from tor-resolve,
but let's avoid linking openssl into tor-resolve for as long as we
can.
This bug shouldn't be reachable so long as secret_to_key_len and
secret_to_key_make_specifier stay in sync, but we might screw up
someday.
Found by coverity; this is CID 1241500
When size_t is the most memory you can have, make sure that things
referring to real parts of memory are size_t, not uint64_t or off_t.
But not on any released Tor.
This implementation allows somebody to add a blinding factor to a
secret key, and a corresponding blinding factor to the public key.
Robert Ransom came up with this idea, I believe. Nick Hopper proved a
scheme like this secure. The bugs are my own.
For proposal 228, we need to cross-certify our identity with our
curve25519 key, so that we can prove at descriptor-generation time
that we own that key. But how can we sign something with a key that
is only for doing Diffie-Hellman? By converting it to the
corresponding ed25519 point.
See the ALL-CAPS warning in the documentation. According to djb
(IIUC), it is safe to use these keys in the ways that ntor and prop228
are using them, but it might not be safe if we start providing crazy
oracle access.
(Unit tests included. What kind of a monster do you take me for?)
This is another case where DJB likes sticking the whole signature
prepended to the message, and I don't think that's the hottest idea.
The unit tests still pass.
This reduces the likelihood that I have made any exploitable errors
in the encoding/decoding.
This commit also imports the trunnel runtime source into Tor.
Uses libscrypt when found; otherwise, we don't have scrypt and we
only support openpgp rfc2440 s2k hashing, or pbkdf2.
Includes documentation and unit tests; coverage around 95%. Remaining
uncovered code is sanity-checks that shouldn't be reachable fwict.
Since address.c is the first file to get compiled, let's have it use
a little judicious c99 in order to catch broken compilers that
somehow make it past our autoconf tests.
"The NULL pointer warnings on the return value of
tor_addr_to_in6_addr32() are incorrect. But clang can't work this
out itself due to limited analysis depth. To teach the analyser that
the return value is safe to dereference, I applied tor_assert to the
return value."
Patch from teor. Part of 13157.
(And replay them once we know our first real logs.)
This is an implementation for issue 6938. It solves the problem of
early log mesages not getting sent to log files, but not the issue of
early log messages not getting sent to controllers.
This fixes bug 13102 (not on any released Tor) where using the
standard SSIZE_MAX name broke mingw64, and we didn't realize.
I did this with
perl -i -pe 's/SIZE_T_MAX/SIZE_MAX/' src/*/*.[ch] src/*/*/*.[ch]
Otherwise, when we're out of input *and* finalizing, we might report
TOR_ZLIB_OK erroneously and not finalize the buffer.
(I don't believe this can happen in practice, with our code today:
write_to_buf_zlib ensures that we are never trying to write into a
completely empty buffer, and zlib says "Z_OK" if you give it even
one byte to write into.)
Fixes bug 11824; bugfix on 0.1.1.23 (06e09cdd47).
We're calling mallocfn() and reallocfn() in the HT_GENERATE macro
with the result of a product. But that makes any sane analyzer
worry about overflow.
This patch keeps HT_GENERATE having its old semantics, since we
aren't the only project using ht.h. Instead, define a HT_GENERATE2
that takes a reallocarrayfn.
Most of these are in somewhat non-obvious code where it is probably
a good idea to initialize variables and add extra assertions anyway.
Closes 13036. Patches from "teor".
This commit attempts to satisfy nickm's comment on check_private_dir() permissions:
https://trac.torproject.org/projects/tor/ticket/11291#comment:12
"""check_private_dir() ensures that the directory has bits 0700 if CPD_CHECK_MODE_ONLY is not set. Shouldn't it also ensure that the directory has bits 0050 if CPD_CHECK_MODE_ONLY is not set, and CPD_GROUP_READ is set?"""
Any error when acquiring the CryptoAPI context should get treated as
bad. Also, this one can't happen for the arguments we're giving.
Fixes bug 10816; bugfix on 0.0.2pre26.
Previously, we had done this only in the connection_free() case, but
when we called connection_free_() directly from
connections_free_all(), we didn't free the connections.
Using the *_array() functions here confused coverity, and was actually
a bit longer than we needed. Now we just use macros for the repeated
bits, so that we can mention a file and a suffix-appended version in
one line.
Previously, we had documented it to return -1 or 0, when in fact
lseek returns -1 or the new position in the file.
This is harmless, since we were only checking for negative values
when we used tor_fd_seekend.
This function is supposed to construct a list of all the ciphers in
the "v2 link protocol cipher list" that are supported by Tor's
openssl. It does this by invoking ssl23_get_cipher_by_char on each
two-byte ciphersuite ID to see which ones give a match. But when
ssl23_get_cipher_by_char cannot find a match for a two-byte SSL3/TLS
ciphersuite ID, it checks to see whether it has a match for a
three-byte SSL2 ciphersuite ID. This was causing a read off the end
of the 'cipherid' array.
This was probably harmless in practice, but we shouldn't be having
any uninitialized reads.
(Using ssl23_get_cipher_by_char in this way is a kludge, but then
again the entire existence of the v2 link protocol is kind of a
kludge. Once Tor 0.2.2 clients are all gone, we can drop this code
entirely.)
Found by starlight. Fix on 0.2.4.8-alpha. Fixes bug 12227.
Found because LibreSSL has OPENSSL_NO_COMP always-on, but this
conflicts with the way that _we_ turn off compression. Patch from
dhill, who attributes it to "OpenBSD". Fixes bug 12602; bugfix on
0.2.1.1-alpha, which introduced this turn-compression-off code.
Long ago we supported systems where there was no support for
threads, or where the threading library was broken. We shouldn't
have do that any more: on every OS that matters, threads exist, and
the OS supports running threads across multiple CPUs.
This resolves tickets 9495 and 12439. It's a prerequisite to making
our workqueue code work better, since sensible workqueue
implementations don't split across multiple processes.
When we create a process yourself with CreateProcess, we get a
handle to the process in the PROCESS_INFO output structure. But
instead of using that handle, we were manually looking up a _new_
handle based on the process ID, which is a poor idea, since the
process ID might refer to a new process later on, but the handle
can't.
This lets us avoid sending SIGTERM to something that has already
died, since we realize it has already died, and is a fix for the
unix version of #8746.
This function is supposed to construct a list of all the ciphers in
the "v2 link protocol cipher list" that are supported by Tor's
openssl. It does this by invoking ssl23_get_cipher_by_char on each
two-byte ciphersuite ID to see which ones give a match. But when
ssl23_get_cipher_by_char cannot find a match for a two-byte SSL3/TLS
ciphersuite ID, it checks to see whether it has a match for a
three-byte SSL2 ciphersuite ID. This was causing a read off the end
of the 'cipherid' array.
This was probably harmless in practice, but we shouldn't be having
any uninitialized reads.
(Using ssl23_get_cipher_by_char in this way is a kludge, but then
again the entire existence of the v2 link protocol is kind of a
kludge. Once Tor 0.2.2 clients are all gone, we can drop this code
entirely.)
Found by starlight. Fix on 0.2.4.8-alpha. Fixes bug 12227.
The old cache had problems:
* It needed to be manually preloaded. (It didn't remember any
address you didn't tell it to remember)
* It was AF_INET only.
* It looked at its cache even if the sandbox wasn't turned on.
* It couldn't remember errors.
* It had some memory management problems. (You can't use memcpy
to copy an addrinfo safely; it has pointers in.)
This patch fixes those issues, and moves to a hash table.
Fixes bug 11970; bugfix on 0.2.5.1-alpha.
These are needed under some circumstances if we are running with
expensive-hardening and sandbox at the same time.
fixes 11477, bugfix on 0.2.5.4-alpha (where we introduced
expensive-hardening)
None of the things we might exec() can possibly run under the
sanbox, so rather than crash later, we have to refuse to accept the
configuration nice and early.
The longer-term solution is to have an exec() helper, but wow is
that risky.
fixes 12043; bugfix on 0.2.5.1-alpha
When running with User set, we frequently try to look up our
information in the user database (e.g., /etc/passwd). The seccomp2
sandbox setup doesn't let us open /etc/passwd, and probably
shouldn't.
To fix this, we have a pair of wrappers for getpwnam and getpwuid.
When a real call to getpwnam or getpwuid fails, they fall back to a
cached value, if the uid/gid matches.
(Granting access to /etc/passwd isn't possible with the way we
handle opening files through the sandbox. It's not desirable either.)
On OpenBSD 5.4, time_t is a 32-bit integer. These instances contain
implicit treatment of long and time_t as comparable types, so explicitly
cast to time_t.
The memarea_strndup() function would have hit undefined behavior by
creating an 'end' pointer off the end of a string if it had ever been
given an 'n' argument bigger than the length of the memory ares that
it's scanning. Fortunately, we never did that except in the unit
tests. But it's not a safe behavior to leave lying around.
If we had an address of the form "1.2.3.4" and we tried to pass it to
tor_inet_pton with AF_INET6, it was possible for our 'eow' pointer to
briefly move backwards to the point before the start of the string,
before we moved it right back to the start of the string. C doesn't
allow that, and though we haven't yet hit a compiler that decided to
nuke us in response, it's best to fix.
So, be more explicit about requiring there to be a : before any IPv4
address part of the IPv6 address. We would have rejected addresses
without a : for not being IPv6 later on anyway.
Previously we said "Sandbox is not implemented on this platform" on
Linux boxes without libseccomp. Now we say that you need to build
Tor built with libseccomp. Fixes bug 11543; bugfix on 0.2.5.1-alpha.
The server cipher list is (thanks to #11513) chosen systematically to
put the best choices for Tor first. The client cipher list is chosen
to resemble a browser. So let's set SSL_OP_CIPHER_SERVER_PREFERENCE
to have the servers pick according to their own preference order.
Older versions of Libevent are happy to open SOCK_DGRAM sockets
non-cloexec and non-nonblocking, and then set those flags
afterwards. It's nice to be able to allow a flag to be on or off in
the sandbox without having to enumerate all its values.
Also, permit PF_INET6 sockets. (D'oh!)
Libevent uses an arc4random implementation (I know, I know) to
generate DNS transaction IDs and capitalization. But it liked to
initialize it either with opening /dev/urandom (which won't work
under the sandbox if it doesn't use the right pointer), or with
sysctl({CTL_KERN,KERN_RANDOM,RANDOM_UUIC}). To make _that_ work, we
were permitting sysctl unconditionally. That's not such a great
idea.
Instead, we try to initialize the libevent PRNG _before_ installing
the sandbox, and make sysctl always fail with EPERM under the
sandbox.
The compiler doesn't warn about this code:
rc = seccomp_rule_add(ctx, SCMP_ACT_ALLOW, SCMP_SYS(openat), 1,
SCMP_CMP(0, SCMP_CMP_EQ, AT_FDCWD),
SCMP_CMP(1, SCMP_CMP_EQ, param->value),
SCMP_CMP(2, SCMP_CMP_EQ, O_RDONLY|...));
but note that the arg_cnt argument above is only 1. This means that
only the first filter (argument 0 == AT_FDCWD) is actually checked!
This patch also fixes the above error in the openat() filter.
Earlier I fixed corresponding errors in filters for rename() and
mprotect().
Appearently, the majority of the filenames we pass to
sandbox_cfg_allow() functions are "freeable right after". So, consider
_all_ of them safe-to-steal, and add a tor_strdup() in the few cases
that aren't.
(Maybe buggy; revise when I can test.)
(If we don't restrict rename, there's not much point in restricting
open, since an attacker could always use rename to make us open
whatever they want.)
A new set of unit test cases are provided, as well as introducing
an alternative paradigm and macros to support it. Primarily, each test
case is given its own namespace, in order to isolate tests from each
other. We do this by in the usual fashion, by appending module and
submodule names to our symbols. New macros assist by reducing friction
for this and other tasks, like overriding a function in the global
namespace with one in the current namespace, or declaring integer
variables to assist tracking how many times a mock has been called.
A set of tests for a small-scale module has been included in this
commit, in order to highlight how the paradigm can be used. This
suite gives 100% coverage to status.c in test execution.
Back in 175b2678, we allowed servers to recognize clients who are
telling them the truth about their ciphersuites, and select the best
cipher from on that list. This implemented the server side of proposal
198.
In bugs 11492, 11498, and 11499, cypherpunks found a bunch of mistakes
and omissions and typos in the UNRESTRICTED_SERVER_CIPHER_LIST we had.
In #11513, I found a couple more.
Rather than try to hand-edit this list, I wrote a short python script
to generate our ciphersuite preferences from the openssl headers.
The new rules are:
* Require forward secrecy.
* Require RSA (since our servers only configure RSA keys)
* Require AES or 3DES. (This means, reject RC4, DES, SEED, CAMELLIA,
and NULL.)
* No export ciphersuites.
Then:
* Prefer AES to 3DES.
* If both suites have the same cipher, prefer ECDHE to DHE.
* If both suites have the same DHE group type, prefer GCM to CBC.
* If both suites have the same cipher mode, prefer SHA384 to SHA256
to SHA1.
* If both suites have the same digest, prefer AES256 to AES128.
This involves some duplicate code between backtrace.c and sandbox.c,
but I don't see a way around it: calling more functions would mean
adding more steps to our call stack, and running clean_backtrace()
against the wrong point on the stack.
The major changes are to re-order some ciphers, to drop the ECDH suites
(note: *not* ECDHE: ECDHE is still there), to kill off some made-up
stuff (like the SSL_RSA_FIPS_WITH_3DES_EDE_CBC_SHA suite), to drop
some of the DSS suites... *and* to enable the ECDHE+GCM ciphersuites.
This change is autogenerated by get_mozilla_ciphers.py from
Firefox 28 and OpenSSL 1.0.1g.
Resolves ticket 11438.
Most of these are simple. The only nontrivial part is that our
pattern for using ENUM_BF was confusing doxygen by making declarations
that didn't look like declarations.
ubsan doesn't like us to do (1u<<32) when 32 is wider than
unsigned. Fortunately, we already special-case
addr_mask_get_bits(0), so we can just change the loop bounds.
This make clang's memory sanitizer happier that we aren't reading
off the end of a char[1]. We hadn't replaced the char[1] with a
char[FLEXIBLE_ARRAY_MEMBER] before because we were doing a union
trick to force alignment. Now we use __attribute__(aligned) where
available, and we do the union trick elsewhere.
Most of this patch is just replacing accesses to (x)->u.mem with
(x)->U_MEM, where U_MEM is defined as "u.mem" or "mem" depending on
our implementation.
This contains the obvious implementation using the circuitmux data
structure. It also runs the old (slow) algorithm and compares
the results of the two to make sure that they're the same.
Needs review and testing.
In a couple of places, to implement the OOM-circuit-killer defense
against sniper attacks, we have counters to remember the age of
cells or data chunks. These timers were based on wall clock time,
which can move backwards, thus giving roll-over results for our age
calculation. This commit creates a low-budget monotonic time, based
on ratcheting gettimeofday(), so that even in the event of a time
rollback, we don't do anything _really_ stupid.
A future version of Tor should update this function to do something
even less stupid here, like employ clock_gettime() or its kin.
It's possible for two threads to hit assertion failures at the same
time. If that happens, let's keep them from stomping on the same
cb_buf field.
Fixes bug 11048; bugfix on 0.2.5.2-alpha. Reported by "cypherpunks".
clang 3.4 introduced a new by-default warning about unused static
functions, which we triggered heavily for the hashtable and map function
generating macros. We can use __attribute__ ((unused)) (thanks nickm for
the suggestion :-) ) to silence these warnings.
It's increasingly apparent that we want to make sure we initialize our
PRNG nice and early, or else OpenSSL will do it for us. (OpenSSL
doesn't do _too_ bad a job, but it's nice to do it ourselves.)
We'll also need this for making sure we initialize the siphash key
before we do any hashes.
I've made an exception for cases where I'm sure that users can't
influence the inputs. This is likely to cause a slowdown somewhere,
but it's safer to siphash everything and *then* look for cases to
optimize.
This patch doesn't actually get us any _benefit_ from siphash yet,
since we don't really randomize the key at any point.
This time, we use a pthread_attr to make sure that if pthread_create
succeeds, the thread is successfully detached.
This probably isn't the big thing going on with 4345, since it'd be
a bit weird for pthread_detach to be failing. But it's worth
getting it right.
This fixes bug 10402, where the rdrand engine would use the rdrand
instruction, not as an additional entropy source, but as a replacement
for the entire userspace PRNG. That's obviously stupid: even if you
don't think that RDRAND is a likely security risk, the right response
to an alleged new alleged entropy source is never to throw away all
previously used entropy sources.
Thanks to coderman and rl1987 for diagnosing and tracking this down.
This flag prevents the creation of a console window popup on Windows. We
need it for pluggable transport executables--otherwise you get blank
console windows when you launch the 3.x browser bundle with transports
enabled.
http://msdn.microsoft.com/en-us/library/ms684863.aspx#CREATE_NO_WINDOW
The browser bundles that used Vidalia used to set this flag when
launching tor itself; it was apparently inherited by the pluggable
transports launched by tor. In the 3.x bundles, tor is launched by some
JavaScript code, which doesn't have the ability to set CREATE_NO_WINDOW.
tor itself is now being compiled with the -mwindows option, so that it
is a GUI application, not a console application, and doesn't show a
console window in any case. This workaround doesn't work for pluggable
transports, because they need to be able to write control messages to
stdout.
https://trac.torproject.org/projects/tor/ticket/9444#comment:30
The previous commit from piet would have backed out some of proposal
198 and made servers built without the V2 handshake not use the
unrestricted cipher list from prop198.
Bug not in any released Tor.
We had accidentially grown two fake ones: one for backtrace.c, and one
for sandbox.c. Let's do this properly instead.
Now, when we configure logs, we keep track of fds that should get told
about bad stuff happening from signal handlers. There's another entry
point for these that avoids using non-signal-handler-safe functions.
On platforms with the backtrace/backtrace_symbols_fd interface, Tor
can now dump stack traces on assertion failure. By default, I log
them to DataDir/stack_dump and to stderr.
Conflicts:
src/or/or.h
src/or/relay.c
Conflicts were simple to resolve. More fixes were needed for
compilation, including: reinstating the tv_to_msec function, and renaming
*_conn_cells to *_chan_cells.
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.
This is probably not an exploitable bug, since you would need to have
errno be a large negative value in the unix pluggable-transport launcher
case. Still, best avoided.
Fixes bug 9928; bugfix on 0.2.3.18-rc.
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.
tor_malloc returns void *; in C, it is not necessary to cast a
void* to another pointer type before assigning it.
tor_malloc fails with an error rather than returning NULL; it's not
necessary to check its output. (In one case, doing so annoyed Coverity.)
SCMP_CMP(a,b,c) leaves the fourth field of the structure undefined,
giving a missing-initializer error. All of our uses are
three-argument, so I'm overriding the default.
Incidentally, this business here where I make crypto_rand mockable:
this is exactly the kind of thing that would make me never want to
include test-support stuff in production builds.
This way, we don't have to use snprintf, which is not guaranteed to
be signal-safe.
(Technically speaking, strlen() and strlcpy() are not guaranteed to
be signal-safe by the POSIX standard. But I claim that they are on
every platform that supports libseccomp2, which is what matters
here.)
Better tests for upper bounds, and for failing cases.
Also, change the function's interface to take a buffer length rather
than a maximum length, and then NUL-terminate: functions that don't
NUL-terminate are trouble waiting to happen.
The only thing that used format_helper_exit_status on win32 was the
unit tests. This caused an error when we tried to leave a static
format_helper_exit_status lying around in a production object file.
The easiest solution is to admit that this way of dealing with process
exit status is Unix-only.
This is not the most beautiful possible implementation (it requires
decorating mockable functions with ugly macros), but it actually
works, and is portable across multiple compilers and architectures.
If you pass the --enable-coverage flag on the command line, we build
our testing binaries with appropriate options eo enable coverage
testing. We also build a "tor-cov" binary that has coverage enabled,
for integration tests.
On recent OSX versions, test coverage only works with clang, not gcc.
So we warn about that.
Also add a contrib/coverage script to actually run gcov with the
appropriate options to generate useful .gcov files. (Thanks to
automake, the .o files will not have the names that gcov expects to
find.)
Also, remove generated gcda and gcno files on clean.
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 mainly a matter of automake trickery: we build each static
library in two versions now: one with the TOR_UNIT_TESTS macro
defined, and one without. When TOR_UNIT_TESTS is defined, we can
enable mocking and expose more functions. When it's not defined, we
can lock the binary down more.
The alternatives would be to have alternate build modes: a "testing
configuration" for building the libraries with test support, and a
"production configuration" for building them without. I don't favor
that approach, since I think it would mean more people runnning
binaries build for testing, or more people not running unit tests.
This implements "algorithm 1" from my discussion of bug #9072: on OOM,
find the circuits with the longest queues, and kill them. It's also a
fix for #9063 -- without the side-effects of bug #9072.
The memory bounds aren't perfect here, and you need to be sure to
allow some slack for the rest of Tor's usage.
This isn't a perfect fix; the rest of the solutions I describe on
codeable.
This reverts commit 884a0e269c.
I'm reverting this because it doesn't actually make the problem go
away. It appears that instead we need to do unmap-then-replace.
A comment by rransom on #8795 taken together with a comment by doorss
recorded on #2077 suggest that *every* attempt to replace the md cache
will fail on Vista/Win7 if we don't have the FILE_SHARE_DELETE flag
passed to CreateFile, and if we try to replace the file ourselves
before unmapping it. I'm adding the FILE_SHARE_DELETE, since that's
this simplest fix. Broken indexers (the favored #2077 hypothesis)
could still cause trouble here, but at least this patch should make us
stop stepping on our own feet.
Likely fix for #2077 and its numerous duplicates. Bugfix on
0.2.2.6-alpha, which first had a microdescriptor cache that would get
replaced before remapping it.
