When fetching a descriptor, we know test every introduction points in it
against our rend failure cache to know if we keep it or not. For this to
work, now everytime an introduction points is discareded (ex: receiving a
NACK), we note it down in our introduction cache.
See rendcache.c for a detailed explanation of the cache's behavior.
Fixes#16389
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
When we removed Running/Valid checks from Fast and Stable in 8712, I
removed them from HSDir too, which apparently wasn't a good idea.
Reverts part of a65e835800. Fixes bug 16524. Bugfix
on 0.2.7.2-alpha.
microdesc_free_() called get_microdesc_cache(), which had the fun
side-effect of potentially reloading the whole cache from disk.
Replace it with a variant that doesn't.
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.
The only reason 16 byte alignment is required is for SSE2 load and
store operations, so only align datastructures to 16 byte boundaries
when building with SSE2 support.
This fixes builds with GCC SSP on platforms that don't have special
case code to do dynamic stack re-alignment (everything not x86/x86_64).
Fixes bug #16666.
The workqueue test help message has two issues. First, the message uses 4 space
indentation when 2 space indentation seems more common. Second, the help
message misses some options.
This commit fixes both issues.
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
It did a good idea, but the code-quality of libupnpc and libnatpnp
is so dodgy that I'm not really comfortable including them alongside
Tor proper. Instead, we'll recommend that people do the pure-go
reimplementation instead. Closes ticket 13338.
Make sure that signing certs are signed by the right identity key,
to prevent a recurrence of #16530. Also make sure that the master
identity key we find on disk matches the one we have in RAM, if we
have one.
This is for #16581.
When there is a signing key and the certificate lists a key, make
sure that the certificate lists the same signing key.
When there are public key and secret key stored in separate files,
make sure they match.
Use the right file name when we load an encrypted secret key and
then find a problem with it.
This is part of 16581.
Add a new and slow unit test that checks if libscrypt_scrypt() and
EBP_PBE_scrypt() yield the same keys from test vectors.
squash! Assert interoperability betweeen libscrypt and OpenSSL EBP_PBE_scrypt().
squash! Assert interoperability betweeen libscrypt and OpenSSL EBP_PBE_scrypt().
squash! Assert interoperability betweeen libscrypt and OpenSSL EBP_PBE_scrypt().
When cleaning up extra circuits that we've opened for performance reason, we
need to count all the introduction circuit and not only the established ones
else we can end up with too many introduction points.
This also adds the check for expiring nodes when serving an INTRODUCE cell
since it's possible old clients are still using them before we have time to
close them.
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
To upload a HS descriptor, this commits makes it that we wait for all
introduction point to be fully established.
Else, the HS ends up uploading a descriptor that may contain intro points
that are not yet "valid" meaning not yet established or proven to work. It
could also trigger three uploads for the *same* descriptor if every intro
points takes more than 30 seconds to establish because of desc_is_dirty
being set at each intro established.
To achieve that, n_intro_points_established varialbe is added to the
rend_service_t object that is incremented when we established introduction
point and decremented when we remove a valid intro point from our list.
The condition to upload a descriptor also changes to test if all intro
points are ready by making sure we have equal or more wanted intro points
that are ready.
The desc_id_dirty flag is kept to be able to still use the
RendInitialPostPeriod option.
This partially fixes#13483.
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
There is a case where if the introduction circuit fails but the node is
still in the consensus, we clean up the intro point and choose an other one.
This commit fixes that by trying to reuse the existing intro point with a
maximum value of retry.
A retry_nodes list is added to rend_services_introduce() and when we remove
an invalid intro points that fits the use case mentionned before, we add the
node to the retry list instead of removing it. Then, we retry on them before
creating new ones.
This means that the requirement to remove an intro point changes from "if no
intro circuit" to "if no intro circuit then if no node OR we've reached our
maximum circuit creation count".
For now, the maximum retries is set to 3 which it completely arbitrary. It
should also at some point be tied to the work done on detecting if our
network is down or not.
Fixes#8239
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
The reasoning for refactoring this function is that removing the
introduction point adaptative algorithm (#4862) ended up changing quite a
bit rend_services_introduce(). Also, to fix some open issues (#8239, #8864
and #13483), this work had to be done.
First, this removes time_expiring variable in an intro point object and
INTRO_POINT_EXPIRATION_GRACE_PERIOD trickery and use an expiring_nodes list
where intro nodes that should expire are moved to that list and cleaned up
only once the new descriptor is successfully uploaded. The previous scheme
was adding complexity and arbitrary timing to when we expire an intro point.
We keep the intro points until we are sure that the new descriptor is
uploaded and thus ready to be used by clients. For this,
rend_service_desc_has_uploaded() is added to notify the HS subsystem that
the descriptor has been successfully uploaded. The purpose of this function
is to cleanup the expiring nodes and circuits if any.
Secondly, this adds the remove_invalid_intro_points() function in order to
split up rend_services_introduce() a bit with an extra modification to it
that fixes#8864. We do NOT close the circuit nor delete the intro point if
the circuit is still alive but the node was removed from the consensus. Due
to possible information leak, we let the circuit and intro point object
expire instead.
Finally, the whole code flow is simplified and large amount of documentation
has been added to mostly explain the why of things in there.
Fixes#8864
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
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.
This needs to be done to allow for the possibility of removing the
ref10 code at a later date, though it is not performance critical.
When integrated by kludging it into tor, it passes unit tests, and is
twice as fast.
Integrating it the "wrong" way into common/crypto_ed25519.c passes
`make check`, and there appear to be some known answer tests for this,
so I assume I got it right.
Blinding a public key goes from 139.10 usec to 70.78 usec using
ed25519-donna (NB: Turboboost/phase of moon), though the code isn't
critical path, so supporting it is mostly done for completeness.
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.
The following arguments change how chutney verifies the network:
--bytes n sends n bytes per test connection (10 KBytes)
--connections n makes n test connections per client (1)
--hs-multi-client 1 makes each client connect to each HS (0)
Requires the corresponding chutney performance testing changes.
Note: using --connections 7 or greater on a HS will trigger #15937.
Patch by "teor".
This is a way to specify the amount of introduction points an hidden service
can have. Maximum value is 10 and the default is 3.
Fixes#4862
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
When we ran out of intro points for a hidden service (which could
happen on a newnym), we would change the connection's state back to
"waiting for hidden service descriptor." But this would make an
assertion fail if we went on to call circuit_get_open_circ_or_launch
again.
This fixes bug 16013; I believe the bug was introduced in
38be533c69, where we made it possible for
circuit_get_open_circ_or_launch() to change the connection's state.
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.
When I fixed#11243, I made it so we would take the digest of a
descriptor before tokenizing it, so we could desist from download
attempts if parsing failed. But when I did that, I didn't remove an
assertion that the descriptor began with "onion-key". Usually, this
was enforced by "find_start_of_next_microdescriptor", but when
find_start_of_next_microdescriptor returned NULL, the assertion was
triggered.
Fixes bug 16400. Thanks to torkeln for reporting and
cypherpunks_backup for diagnosing and writing the first fix here.
Every functions and objects that are used for hidden service descriptor
caches are moved to rendcache.{c|h}.
This commit does NOT change anything, just moving code around.
Fixes#16399
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
For now, rend_cache_entry_t has been moved from or.h to rendcache.h and
those files have been added to the build system.
In the next commit, these will contain hidden service descriptor cache ABI
and API for both client and directory side. The goal is to consolidate the
descriptor caches in one location to ease development, maintenance, review
and improve documentation for each cache behavior and algorithm.
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
When --keygen is provided, we prompt for a passphrase when we make a
new master key; if it is nonempty, we store the secret key in a new
crypto_pwbox.
Also, if --keygen is provided and there *is* an encrypted master key,
we load it and prompt for a passphrase unconditionally.
We make a new signing key unconditionally when --keygen is provided.
We never overwrite a master key.
If crypto_early_init fails, a typo in a return value from tor_init
means that tor_main continues running, rather than returning
an error value.
Fixes bug 16360; bugfix on d3fb846d8c in 0.2.5.2-alpha,
introduced when implementing #4900.
Patch by "teor".
This reverts commit 9407040c59.
Small fix, "e->received" had to be removed since that variable doesn't exist
anymore.
Signed-off-by: David Goulet <dgoulet@ev0ke.net>
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.
signing_key can be NULL in ed_key_init_from_file in routerkeys.c.
Discovered by clang 3.7 address sanitizer.
Fix on c03694938e, not in any released version of Tor.
clang 3.7 complains that using a preprocessor directive inside
a macro invocation in test_util_writepid in test_util.c is undefined.
Fix on 79e85313aa on 0.2.7.1-alpha.
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.
Rend_add_service() frees its argument on failure; no need to free again.
Fixes bug 16228, bugfix on 0.2.7.1-alpha
Found by coverity; this is CID 1301387.
Also, have testing-level options to set the lifetimes and
expiration-tolerances of all key types, plus a non-testing-level
option to set the lifetime of any auto-generated signing key.
# The first commit's message is:
Regenerate ed25519 keys when they will expire soon.
Also, have testing-level options to set the lifetimes and
expiration-tolerances of all key types, plus a non-testing-level
option to set the lifetime of any auto-generated signing key.
# The 2nd commit message will be skipped:
# fixup! Regenerate ed25519 keys when they will expire soon.
This is a new collator type that follows proposal 220 for deciding
which identities to include. The rule is (approximately):
If a <ed,rsa> identity is listed by more than half of authorities,
include it. And include all <rsa> votes about that node as
matching.
Otherwise, if an <*,rsa> or <rsa> identity is listed by more than
half of the authorities, and no <ed,rsa> has been listed, include
it.
* Include ed25519 identities in votes
* Include "no ed25519 identity" in votes
* Include some commented-out code about identity voting. (This
will disappear.)
* Include some functions for identity voting (These will disappear.)
* Enforce uniqueness in ed25519 keys within a vote
Extrainfo documents are now ed-signed just as are router
descriptors, according to proposal 220. This patch also includes
some more tests for successful/failing parsing, and fixes a crash
bug in ed25519 descriptor parsing.
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.
This includes the link handshake variations for proposal220.
We'll use this for testing first, and then use it to extend our
current code to support prop220.
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.
With this patch:
* Authorities load the key-pinning log at startup.
* Authorities open a key-pinning log for writing at startup.
* Authorities reject any router with an ed25519 key where they have
previously seen that ed25519 key with a different RSA key, or vice
versa.
* Authorities warn about, but *do not* reject, RSA-only descriptors
when the RSA key has previously gone along with an Ed25519 key.
(We should make this a 'reject' too, but we can't do that until we're
sure there's no legit reason to downgrade to 0.2.5.)
This module implements a key-pinning mechanism to ensure that it's
safe to use RSA keys as identitifers even as we migrate to Ed25519
keys. It remembers, for every Ed25519 key we've seen, what the
associated Ed25519 key is. This way, if we see a different Ed25519
key with that RSA key, we'll know that there's a mismatch.
We persist these entries to disk using a simple format, where each
line has a base64-encoded RSA SHA1 hash, then a base64-endoded
Ed25519 key. Empty lines, misformed lines, and lines beginning with
a # are ignored. Lines beginning with @ are reserved for future
extensions.
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.
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.
