Our convention is that functions which manipulate a type T should be
named T_foo. But the buffer functions were super old, and followed
all kinds of conventions. Now they're uniform.
Here's the perl I used to do this:
\#!/usr/bin/perl -w -i -p
s/read_to_buf\(/buf_read_from_socket\(/;
s/flush_buf\(/buf_flush_to_socket\(/;
s/read_to_buf_tls\(/buf_read_from_tls\(/;
s/flush_buf_tls\(/buf_flush_to_tls\(/;
s/write_to_buf\(/buf_add\(/;
s/write_to_buf_compress\(/buf_add_compress\(/;
s/move_buf_to_buf\(/buf_move_to_buf\(/;
s/peek_from_buf\(/buf_peek\(/;
s/fetch_from_buf\(/buf_get_bytes\(/;
s/fetch_from_buf_line\(/buf_get_line\(/;
s/fetch_from_buf_line\(/buf_get_line\(/;
s/buf_remove_from_front\(/buf_drain\(/;
s/peek_buf_startswith\(/buf_peek_startswith\(/;
s/assert_buf_ok\(/buf_assert_ok\(/;
These statistics were largely ununsed, and kept track of statistical information
on things like how many time we had done TLS or how many signatures we had
verified. This information is largely not useful, and would only be logged
after receiving a SIGUSR1 signal (but only if the logging severity level was
less than LOG_INFO).
* FIXES#19871.
* REMOVES note_crypto_pk_op(), dump_pk_op(), and pk_op_counts from
src/or/rephist.c.
* REMOVES every external call to these functions.
This mistake causes two possible bugs. I believe they are both
harmless IRL.
BUG 1: memory stomping
When we call the memset, we are overwriting two 0 bytes past the end
of packed_cell_t.body. But I think that's harmless in practice,
because the definition of packed_cell_t is:
// ...
typedef struct packed_cell_t {
TOR_SIMPLEQ_ENTRY(packed_cell_t) next;
char body[CELL_MAX_NETWORK_SIZE];
uint32_t inserted_time;
} packed_cell_t;
So we will overwrite either two bytes of inserted_time, or two bytes
of padding, depending on how the platform handles alignment.
If we're overwriting padding, that's safe.
If we are overwriting the inserted_time field, that's also safe: In
every case where we call cell_pack() from connection_or.c, we ignore
the inserted_time field. When we call cell_pack() from relay.c, we
don't set or use inserted_time until right after we have called
cell_pack(). SO I believe we're safe in that case too.
BUG 2: memory exposure
The original reason for this memset was to avoid the possibility of
accidentally leaking uninitialized ram to the network. Now
remember, if wide_circ_ids is false on a connection, we shouldn't
actually be sending more than 512 bytes of packed_cell_t.body, so
these two bytes can only leak to the network if there is another bug
somewhere else in the code that sends more data than is correct.
Fortunately, in relay.c, where we allocate packed_cell_t in
packed_cell_new() , we allocate it with tor_malloc_zero(), which
clears the RAM, right before we call cell_pack. So those
packed_cell_t.body bytes can't leak any information.
That leaves the two calls to cell_pack() in connection_or.c, which
use stack-alocated packed_cell_t instances.
In or_handshake_state_record_cell(), we pass the cell's contents to
crypto_digest_add_bytes(). When we do so, we get the number of
bytes to pass using the same setting of wide_circ_ids as we passed
to cell_pack(). So I believe that's safe.
In connection_or_write_cell_to_buf(), we also use the same setting
of wide_circ_ids in both calls. So I believe that's safe too.
I introduced this bug with 1c0e87f6d8
back in 0.2.4.11-alpha; it is bug 22737 and CID 1401591
It's okay to call add_ed25519_cert with a NULL argument: so,
document that. Also, add a tor_assert_nonfatal() to catch any case
where we have failed to set own_link_cert when conn_in_server_mode.
Whenever we rotate our TLS context, we change our Ed25519
Signing->Link certificate. But if we've already started a TLS
connection, then we've already sent the old X509 link certificate,
so the new Ed25519 Signing->Link certificate won't match it.
To fix this, we now store a copy of the Signing->Link certificate
when we initialize the handshake state, and send that certificate
as part of our CERTS cell.
Fixes one case of bug22460; bugfix on 0.3.0.1-alpha.
Accomplished via the following:
1. Use NETINFO cells to determine if both peers will agree on canonical
status. Prefer connections where they agree to those where they do not.
2. Alter channel_is_better() to prefer older orconns in the case of multiple
canonical connections, and use the orconn with more circuits on it in case
of age ties.
Also perform some hourly accounting on how many of these types of connections
there are and log it at info or notice level.
This unifies CircuitIdleTimeout and PredictedCircsRelevanceTime into a single
option, and randomizes it.
It also gives us control over the default value as well as relay-to-relay
connection lifespan through the consensus.
Conflicts:
src/or/circuituse.c
src/or/config.c
src/or/main.c
src/test/testing_common.c
This defense will cause Cisco, Juniper, Fortinet, and other routers operating
in the default configuration to collapse netflow records that would normally
be split due to the 15 second flow idle timeout.
Collapsing these records should greatly reduce the utility of default netflow
data for correlation attacks, since all client-side records should become 30
minute chunks of total bytes sent/received, rather than creating multiple
separate records for every webpage load/ssh command interaction/XMPP chat/whatever
else happens to be inactive for more than 15 seconds.
The defense adds consensus parameters to govern the range of timeout values
for sending padding packets, as well as for keeping connections open.
The defense only sends padding when connections are otherwise inactive, and it
does not pad connections used solely for directory traffic at all. By default
it also doesn't pad inter-relay connections.
Statistics on the total padding in the last 24 hours are exported to the
extra-info descriptors.
If a node can prove its Ed25519 identity, don't consider connections
to it canonical unless they match both identities.
Includes link handshake changes needed to avoid crashing with bug
warnings, since the tests now reach more parts of the code.
Closes ticket 20355
(Only run the connection_or_group_set_badness_() function on groups
of channels that have the same RSA and Ed25519 identities.)
There's a possible opportunity here where we might want to set a
channel to "bad" if it has no ed25519 identity and some other
channel has some. Also there's an opportunity to add a warning if
we ever have an Ed mismatch on open connections with the same RSA
ID.
This function has never gotten testing for the case where an
identity had been set, and then got set to something else. Rather
than make it handle those cases, we forbid them.
This patch makes two absolutely critical changes:
- If an ed25519 identity is not as expected when creating a channel,
we call that channel unsuccessful and close it.
- When a client creating a channel or an extend cell for a circuit, we
only include the ed25519 identity if we believe that the node on
the other side supports ed25519 link authentication (from
#15055). Otherwise we will insist on nodes without the right
link protocol authenticating themselves.
- When deciding to extend to another relay, we only upgrade the
extend to extend by ed25519 ID when we know the ed25519 ID _and_
we know that the other side can authenticate.
This patch also tells directory servers, when probing nodes, to
try to check their ed25519 identities too (if they can authenticate
by ed25519 identity).
Also, handle the case where we connect by RSA Id, and learn the
ED25519 ID for the node in doing so.
Right now, there's only a mechanism to look for a channel where the
RSA ID matches *and* the ED ID matches. We can add a separate map
later if we want.
This patch doesn't cover every case; omitted cases are marked with
"XXXX prop271", as usual. It leaves both the old interface and the
new interface for guard status notification, since they don't
actually work in the same way: the new API wants to be told when a
circuit has failed or succeeded, whereas the old API wants to know
when a channel has failed or succeeded.
I ran into some trouble with directory guard stuff, since when we
pick the directory guard, we don't actually have a circuit to
associate it with. I solved that by allowing guard states to be
associated with directory connections, not just circuits.
This patch is just:
* Code movement
* Adding headers here and there as needed
* Adding a bridges_free_all() with a call to it.
It breaks compilation, since the bridge code needed to make exactly
2 calls into entrynodes.c internals. I'll fix those in the next
commit.
