The variable was useless since it was only toggled off in disabled code.
If the 'exit_family' smartlist is empty, we don't consider exit family
anyway.
The fix for bug 8746 added a hashtable instance that never actually
invoked HT_FIND. This caused a warning, since we didn't mark HT_FIND
as okay-not-to-use.
Try killing a running process; try noticing that a process has
exited without checking its output; verify that waitpid_cb (when
present) is set to NULL when you would expect it to be.
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.
Check for consistency between the queued destroy cells and the marked
circuit IDs. Check for consistency in the count of queued destroy
cells in several ways. Check to see whether any of the marked circuit
IDs have somehow been marked longer than the channel has existed.
And add a comment about why conditions that would cause us to drop a
cell should get checked before actions that would cause us to send a
destroy cell.
Spotted by 'cypherpunks'.
And note that these issues have been present since 0.0.8pre1 (commit
0da256ef), where we added a "shutting down" state, and started
responding to all create cells with DESTROY when shutting down.
Conflicts:
src/or/channel.c
src/or/circuitlist.c
src/or/connection.c
Conflicts involved removal of next_circ_id and addition of
unusable-circid tracking.
The point of the "idle timeout" for connections is to kill the
connection a while after it has no more circuits. But using "last
added a non-padding cell" as a proxy for that is wrong, since if the
last circuit is closed from the other side of the connection, we
will not have sent anything on that connection since well before the
last circuit closed.
This is part of fixing 6799.
When applied to 0.2.5, it is also a fix for 12023.
Instead of killing an or_connection_t that has had no circuits for
the last 3 minutes, give every or_connection_t a randomized timeout,
so that an observer can't so easily infer from the connection close
time the time at which its last circuit closed.
Also, increase the base timeout for canonical connections from 3
minutes to 15 minutes.
Fix for ticket 6799.
When we find a stranded one-hop circuit, log whether it is dirty,
log information about any streams on it, and log information about
connections they might be linked to.
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.
This code mis-handled the case where a circuit got the same circuit
ID in both directions. I found three instances of it in the
codebase, by grepping for [pn]_circ_id.
Because of the issue in command_process_relay_cell(), this would
have made roughly one circuit in a million completely nonfunctional.
Fixes bug 12195.
On some profiles of Andrea's from #11332, I found that a great deal
of time can still be attributed to functions called from
update_router_have_minimum_dir_info(). This is making our
digestmap, tor_memeq, and siphash functions take a much bigger
portion of runtime than they really should.
If we're calling update_router_have_minimum_dir_info() too often,
that's because we're calling router_dir_info_changed() too often.
And it looks like most of the callers of router_dir_info_changed()
are coming as tail-calls from router_set_status() as invoked by
channel_do_open_actions().
But we don't need to call router_dir_info_changed() so much! (I'm
not quite sure we need to call it from here at all, but...) Surely
we don't need to call it from router_set_status when the router's
status has not actually changed.
This patch makes us call router_dir_info_changed() from
router_set_status only when we are changing the router's status.
Fix for bug 12170. This is leftover from our fix back in 273ee3e81
in 0.1.2.1-alpha, where we started caching the value of
update_router_have_minimum_dir_info().
tor_memeq has started to show up on profiles, and this is one of the
most frequent callers of that function, appearing as it does on every
cell handled for entry or exit.
59f9097d5c introduced tor_memneq here;
it went into Tor 0.2.1.31. Fixes part of 12169.
Without this fix, when running with bridges, we would try fetching
directory info far too early, and have up to a 60 second delay if we
started with bridge descriptors available.
Fixes bug 11965. Fix on 0.2.3.6-alpha, arma thinks.
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.
The code was not disambiguating ClientTransportPlugin configured and
not used, and ClientTransportPlugin configured, but in a failed state.
The right thing to do is to undo moving the get_transport_by_addrport()
call back into get_proxy_addrport(), and remove and explicit check for
using a Bridge since by the time the check is made, if a Bridge is
being used, it is PT/proxy-less.
This change allows using Socks4Proxy, Socks5Proxy and HTTPSProxy with
ClientTransportPlugins via the TOR_PT_PROXY extension to the
pluggable transport specification.
This fixes bug #8402.
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 we converted the horrible set of options that previously
controlled "use ORPort or DirPort? Anonymously or Non-anonymouly?" to
a single 'indirection' argument, we missed
directory_post_to_dirservers.
The problematic code was introduced in 5cbeb6080, which went into
0.2.4.3-alpha. This is a fix for bug 11469.
If somebody has configured a client to use a bridge without setting
an identity digest (not recommended), learn the identity digest from
whatever bridge descriptor we have downloaded or have in our cache.
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.
These are actually tests for #311. It appears to me that we didn't
fix#311 properly when we thought we did in 475eb5d6; instead, the
real fix was 05eff35ac6, a few minutes earlier.
Clients should always believe that v3 directory authorities serve
extra-info documents, regardless of whether their server descriptor
contains a "caches-extra-info" line or not.
Fixes part of #11683.
on #9686, gmorehose reports that the 500 MB lower limit is too high
for raspberry pi users.
This is a backport of 647248729f to 0.2.4.
Note that in 0.2.4, the option is called MaxMemInCellQueues.
This was previously satisfied by using a temporary variable, but there
are three other instances in circuitlist.c that gcc is now bothered by,
so now introduce a CONST_TO_ORIGIN_CIRCUIT that takes a const
circuit_t instead.
When clearing a list of tokens, it's important to do token_clear()
on them first, or else any keys they contain will leak. This didn't
leak memory on any of the successful microdescriptor parsing paths,
but it does leak on some failing paths when the failure happens
during tokenization.
Fixes bug 11618; bugfix on 0.2.2.6-alpha.
The python scripts invoked by 'make check' didn't work on python3
before. That was a problem on systems where 'python' is python3.
Fixes bug 11608; bugfix on 0.2.5.2-alpha.
If we can't detect the physical memory, the new default is 8 GB on
64-bit architectures, and 1 GB on 32-bit architectures.
If we *can* detect the physical memory, the new default is
CLAMP(256 MB, phys_mem * 0.75, MAX_DFLT)
where MAX_DFLT is 8 GB on 64-bit architectures and 2 GB on 32-bit
architectures.
You can still override the default by hand. The logic here is simply
trying to choose a lower default value on systems with less than 12 GB
of physical RAM.
Use a per-channel ratelim_t to control the rate at which we report
failures for each channel.
Explain why I picked N=32.
Never return a zero circID.
Thanks to Andrea and to cypherpunks.
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.
Instead of taking the length of a buffer, we were taking the length of
a pointer, so that our debugging log would cover only the first
sizeof(void*) bytes of the client nonce.
scan-build recognizes that in theory there could be a numeric overflow
here.
This can't numeric overflow can't trigger IRL, since in order to fill a
hash table with more than P=402653189 buckets with a reasonable load
factor of 0.5, we'd first have P/2 malloced objects to put in it--- and
each of those would have to take take at least sizeof(void*) worth of
malloc overhead plus sizeof(void*) content, which would run you out of
address space anyway on a 32-bit system.
If 'intro' is NULL in these functions, I'm pretty sure that the
error message must be set before we hit the end. But scan-build
doesn't notice that, and is worried that we'll do a null-pointer
dereference in the last-chance errormsg generation.
As it stands, it relies on the fact that onion_queue_entry_remove
will magically remove each onionskin from the right list. This
patch changes the logic to be more resilient to possible bugs in
onion_queue_entry_remove, and less confusing to static analysis tools.
scan-build doesn't realize that a request can't be timed at the end
unless it's timed at the start, and so it's not possible for us to
be subtracting start from end without start being set.
Nevertheless, let's not confuse it.
When get_proxy_addrport returned PROXY_NONE, it would leave
addr/port unset. This is inconsistent, and could (if we used the
function in a stupid way) lead to undefined behavior. Bugfix on
5b050a9b0, though I don't think it affects tor-as-it-is.
Throughout circuituse, when we log about a circuit, we log its
desired path length from build_state. scan-build is irrationally
concerned that build_state might be NULL.
In circuitmux_detach_all_circuits, we check whether an HT iterator
gives us NULL. That should be impossible for an HT iterator. But
our checking it has confused scan-build (justly) into thinking that
our later use of HT_NEXT_RMV might not be kosher. I'm taking the
coward's route here and strengthening the check. Bugfix on
fd31dd44. (Not a real bug though)
If we fail in circuit_get_by_rend_token_and_purpose because the
circuit has no rend_info, don't try to reference fiends from its
rend_info when logging an error. Bugfix on 8b9a2cb68, which is
going into Tor 0.2.5.4-alpha.
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.
Fixes a possible root cause of 11553 by only making 64 attempts at
most to pick a circuitID. Previously, we would test every possible
circuit ID until we found one or ran out.
This algorithm succeeds probabilistically. As the comment says:
This potentially causes us to give up early if our circuit ID
space is nearly full. If we have N circuit IDs in use, then we
will reject a new circuit with probability (N / max_range) ^
MAX_CIRCID_ATTEMPTS. This means that in practice, a few percent
of our circuit ID capacity will go unused.
The alternative here, though, is to do a linear search over the
whole circuit ID space every time we extend a circuit, which is
not so great either.
This makes new vs old clients distinguishable, so we should try to
batch it with other patches that do that, like 11438.
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.
This means that tor can run without needing to communicate with ioctls
to the firewall, and therefore doesn't need to run with privileges to
open the /dev/pf device node.
A new TransProxyType is added for this purpose, "pf-divert"; if the user
specifies this TransProxyType in their torrc, then the pf device node is
never opened and the connection destination is determined with getsockname
(as per pf(4)). The default behaviour (ie., when TransProxyType is "default"
when using the pf firewall) is still to assume that pf is configured with
rdr-to rules.
This isn't on by default; to get it, you need to set "TransProxyType
ipfw". (The original patch had automatic detection for whether
/dev/pf is present and openable, but that seems marginally fragile.)
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.
My first implementation was broken, since it returned "whether there
is one bridge" rather than "how many bridges."
Also, the implementation for the n_options_out feature in
choose_random_entry_impl was completely broken due to a missing *.
When we successfully create a usable circuit after it previously
timed out for a certain amount of time, we should make sure that
our public IP address hasn't changed and update our descriptor.
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.
In C, it's a bad idea to do this:
char *cp = array;
char *end = array + array_len;
/* .... */
if (cp + 3 >= end) { /* out of bounds */ }
because cp+3 might be more than one off the end of the array, and
you are only allowed to construct pointers to the array elements,
and to an element one past the end. Instead you have to say
if (cp - array + 3 >= array_len) { /* ... */ }
or something like that.
This patch fixes two of these: one in process_versions_cell
introduced in 0.2.0.10-alpha, and one in process_certs_cell
introduced in 0.2.3.6-alpha. These are both tracked under bug
10363. "bobnomnom" found and reported both. See also 10313.
In our code, this is likely to be a problem as we used it only if we
get a nasty allocator that makes allocations end close to (void*)-1.
But it's best not to have to worry about such things at all, so
let's just fix all of these we can find.
According to reports, most programs degrade somewhat gracefully on
getting no answer for an MX or a CERT for www.example.com, but many
flip out completely on a NOTIMPL error.
Also, treat a QTYPE_ALL query as just asking for an A record.
The real fix here is to implement proposal 219 or something like it.
Fixes bug 10268; bugfix on 0.2.0.1-alpha.
Based on a patch from "epoch".
Found by testing with chutney. The old behavior was "fail an
assertion", which obviously isn't optimal.
Bugfix on 8b9a2cb68b290e550695124d7ef0511225b451d5; bug not in any
released version.
Also, stop accepting the old kind of RESOLVED cells with no TTL
fields; they haven't been sent since 0.1.1.6-alpha.
This patch won't work without the fix to #10468 -- it will break
DNSPorts unless they set the proper ipv4/6 flags on entry_connection_t.
