The function compat_getdelim_ is used for tor_getline if tor is compiled
on a system that lacks getline and getdelim. These systems should be
very rare, considering that getdelim is POSIX.
If this system is further a 32 bit architecture, it is possible to
trigger a double free with huge files.
If bufsiz has been already increased to 2 GB, the next chunk would
be 4 GB in size, which wraps around to 0 due to 32 bit limitations.
A realloc(*buf, 0) could be imagined as "free(*buf); return malloc(0);"
which therefore could return NULL. The code in question considers
that an error, but will keep the value of *buf pointing to already
freed memory.
The caller of tor_getline() would free the pointer again, therefore
leading to a double free.
This code can only be triggered in dirserv_read_measured_bandwidths
with a huge measured bandwith list file on a system that actually
allows to reach 2 GB of space through realloc.
It is not possible to trigger this on Linux with glibc or other major
*BSD systems even on unit tests, because these systems cannot reach
so much memory due to memory fragmentation.
This patch is effectively based on the penetration test report of
cure53 for curl available at https://cure53.de/pentest-report_curl.pdf
and explained under section "CRL-01-007 Double-free in aprintf() via
unsafe size_t multiplication (Medium)".
If the concatenation of connection buffer and the buffer of linked
connection exceeds INT_MAX bytes, then buf_move_to_buf returns -1 as an
error value.
This value is currently casted to size_t (variable n_read) and will
erroneously lead to an increasement of variable "max_to_read".
This in turn can be used to call connection_buf_read_from_socket to
store more data inside the buffer than expected and clogging the
connection buffer.
If the linked connection buffer was able to overflow INT_MAX, the call
of buf_move_to_buf would have previously internally triggered an integer
overflow, corrupting the state of the connection buffer.
Signed-off-by: Tobias Stoeckmann <tobias@stoeckmann.org>
Many buffer functions have a hard limit of INT_MAX for datalen, but
this limitation is not enforced in all functions:
- buf_move_all may exceed that limit with too many chunks
- buf_move_to_buf exceeds that limit with invalid buf_flushlen argument
- buf_new_with_data may exceed that limit (unit tests only)
This patch adds some annotations in some buf_pos_t functions to
guarantee that no out of boundary access could occur even if another
function lacks safe guards against datalen overflows.
Signed-off-by: Tobias Stoeckmann <tobias@stoeckmann.org>
If the concatenation of connection buffer and the buffer of linked
connection exceeds INT_MAX bytes, then buf_move_to_buf returns -1 as an
error value.
This value is currently casted to size_t (variable n_read) and will
erroneously lead to an increasement of variable "max_to_read".
This in turn can be used to call connection_buf_read_from_socket to
store more data inside the buffer than expected and clogging the
connection buffer.
If the linked connection buffer was able to overflow INT_MAX, the call
of buf_move_to_buf would have previously internally triggered an integer
overflow, corrupting the state of the connection buffer.
Signed-off-by: Tobias Stoeckmann <tobias@stoeckmann.org>
Many buffer functions have a hard limit of INT_MAX for datalen, but
this limitation is not enforced in all functions:
- buf_move_all may exceed that limit with too many chunks
- buf_move_to_buf exceeds that limit with invalid buf_flushlen argument
- buf_new_with_data may exceed that limit (unit tests only)
This patch adds some annotations in some buf_pos_t functions to
guarantee that no out of boundary access could occur even if another
function lacks safe guards against datalen overflows.
[This is a backport of the submitted patch to 0.2.9, where the
buf_move_to_buf and buf_new_with_data functions did not exist.]
Fixes bug 29922; bugfix on 0.2.9.3-alpha when we tried to capture
all these warnings. No need to backport any farther than 0.3.5,
though -- these warnings don't cause test failures before then.
This one was tricky to find because apparently it only happened on
_some_ windows builds.
In current NSS versions, these ciphersuites don't work with
SSL_ExportKeyingMaterial(), which was causing relays to fail when
they tried to negotiate the v3 link protocol authentication.
Fixes bug 29241; bugfix on 0.4.0.1-alpha.
When classifying a client's selection of TLS ciphers, if the client
ciphers are not yet available, do not cache the result. Previously,
we had cached the unavailability of the cipher list and never looked
again, which in turn led us to assume that the client only supported
the ancient V1 link protocol. This, in turn, was causing Stem
integration tests to stall in some cases. Fixes bug 30021; bugfix
on 0.2.4.8-alpha.
Let's use the same function exit point for BUG() codepath that we're using
for every other exit condition. That way, we're not forgetting to clean up
the memarea.
so that the relays that would be "excluded" from the bandwidth
file because of something failed can be included to diagnose what
failed, without still including these relays in the bandwidth
authorities vote.
Closes#29806.
Allow connections to single onion services to remain idle without being
disconnected.
Relays acting as rendezvous points for single onion services were
mistakenly closing idle established rendezvous circuits after 60 seconds,
thinking that they are unused directory-fetching circuits that had served
their purpose.
Fixes bug 29665; bugfix on 0.2.1.26.
Allow connections to single onion services to remain idle without being
disconnected.
Relays acting as rendezvous points for single onion services were
mistakenly closing idle established rendezvous circuits after 60 seconds,
thinking that they are unused directory-fetching circuits that had served
their purpose.
Fixes bug 29665; bugfix on 0.2.1.26.
Stop leaking parts of the shared random state in the shared-random unit
tests. The previous fix in 29599 was incomplete.
Fixes bug 29706; bugfix on 0.2.9.1-alpha.
Turns out that when reloading a tor configured with hidden service(s), we
weren't copying all the needed information between the old service object to
the new one.
For instance, the desc_is_dirty timestamp wasn't which could lead to the
service uploading its descriptor much later than it would need to.
The replaycache wasn't also moved over and some intro point information as
well.
Fixes#23790
Signed-off-by: David Goulet <dgoulet@torproject.org>
Prior to #23100, we were not counting HS circuit build times in our
calculation of the timeout. This could lead to a condition where our timeout
was set too low, based on non HS circuit build times, and then we would
abandon all HS circuits, storing no valid timeouts in the histogram.
This commit avoids the assert.
When "auto" was used for the port number for a listening socket, the
message logged after opening the socket would incorrectly say port 0
instead of the actual port used.
Fixes bug 29144; bugfix on 0.3.5.1-alpha
Signed-off-by: Kris Katterjohn <katterjohn@gmail.com>
Fixes bug 29530, where the LOG_ERR messages were occurring when
we had no configured network, and so we were failing the unit tests
because of the recently-merged #28668.
Commit message edited by teor:
We backported 28668 and released it in 0.3.5.8.
This commit backports 29530 to 0.3.5.
Fixes bug 29530 in Tor 0.3.5.8.
When IPv4Only (IPv6Only) was used but the address could not be
interpreted as a IPv4 (IPv6) address, the error message referred
to the wrong IP version.
This also fixes up the error-checking logic so it's more precise
about what's being checked.
Fixes bug 13221; bugfix on 0.2.3.9-alpha
Signed-off-by: Kris Katterjohn <katterjohn@gmail.com>
KIST works by computing how much should be allowed to write to the kernel for
a given socket, and then it writes that amount to the outbuf.
The problem is that it could be possible that the outbuf already has lots of
data in it from a previous scheduling round (because the kernel is full/busy
and Tor was not able to flush the outbuf yet). KIST ignores that the outbuf
has been filling (is above its "highwater") and writes more anyway. The end
result is that the outbuf length would exceed INT_MAX, hence causing an
assertion error and a corresponding "Bug()" message to get printed to the
logs.
This commit makes it for KIST to take into account the outbuf length when
computing the available space.
Bug found and patch by Rob Jansen.
Closes#29168. TROVE-2019-001.
Signed-off-by: David Goulet <dgoulet@torproject.org>
malloc_options needs to be declared extern (and declaring it extern
means we need to initialize it separately)
Fixes bug 29145; bugfix on 0.2.9.3-alpha
Signed-off-by: Kris Katterjohn <katterjohn@gmail.com>
Because the test is adding entries to the "rend_cache" directly, the
rend_cache_increment_allocation() was never called which made the
rend_cache_clean() call trigger that underflow warning:
rend_cache/clean: [forking] Nov 29 09:55:04.024 [warn] rend_cache_decrement_allocation(): Bug: Underflow in rend_cache_decrement_allocation (on Tor 0.4.0.0-alpha-dev 2240fe63feb9a8cf)
The test is still good and valid.
Fixes#28660
Signed-off-by: David Goulet <dgoulet@torproject.org>
Because the test is adding entries to the "rend_cache" directly, the
rend_cache_increment_allocation() was never called which made the
rend_cache_clean() call trigger that underflow warning:
rend_cache/clean: [forking] Nov 29 09:55:04.024 [warn] rend_cache_decrement_allocation(): Bug: Underflow in rend_cache_decrement_allocation (on Tor 0.4.0.0-alpha-dev 2240fe63feb9a8cf)
The test is still good and valid.
Fixes#28660
Signed-off-by: David Goulet <dgoulet@torproject.org>
The code checked for sysctl being available and HW_PHYSMEM being
defined, but HW_USERMEM was actually being used with sysctl instead
of HW_PHYSMEM.
The case for OpenBSD, etc. use HW_PHYSMEM64 (which is obviously a
64-bit variant of HW_PHYSMEM) and the case for OSX uses HW_MEMSIZE
(which appears to be a 64-bit variant of HW_PHYSMEM).
Signed-off-by: Kris Katterjohn <katterjohn@gmail.com>
We log these messages at INFO level, except when we are reading a
private key from a file, in which case we log at WARN.
This fixes a regression from when we re-wrote our PEM code to be
generic between nss and openssl.
Fixes bug 29042, bugfix on 0.3.5.1-alpha.
Stop logging "Tried to establish rendezvous on non-OR circuit..." as
a warning. Instead, log it as a protocol warning, because there is
nothing that relay operators can do to fix it.
Fixes bug 29029; bugfix on 0.2.5.7-rc.
Prior to this commit, the testsuite was failing on OpenBSD. After
this commit the testsuite runs fine on OpenBSD.
It was previously decided to test for the OpenBSD macro (rather than
__OpenBSD__, etc.) because OpenBSD forks seem to have the former
macro defined. sys/param.h must be included for the OpenBSD macro
definition; however, many files tested for the OpenBSD macro without
having this header included.
This commit includes sys/param.h in the files where the OpenBSD macro
is used (and sys/param.h is not already included), and it also
changes some instances of the __OpenBSD__ macro to OpenBSD.
See commit 27df23abb6 which changed
everything to use OpenBSD instead of __OpenBSD__ or OPENBSD. See
also tickets #6982 and #20980 (the latter ticket is where it was
decided to use the OpenBSD macro).
Signed-off-by: Kris Katterjohn <katterjohn@gmail.com>
In get_local_listener used by tor_ersatz_socketpair, the address
family used when binding the IPv6 socket was AF_INET instead of
AF_INET6.
Fixes bug 28995; bugfix on 0.3.5.1-alpha.
Signed-off-by: Kris Katterjohn <katterjohn@gmail.com>
In theory it would be better to detect this bug in advance, but this
approach is much simpler, and therefore safer to backport.
This closes tor issue 28973.
Linked connections aren't woken up by libevent due to I/O but rather
artificially so we can, by chunks, empty the spooled object(s).
Commit 5719dfb48f (in 0.3.4.1-alpha) made it
that the schedule_active_linked_connections_event would be only called once at
startup but this is wrong because then we would never go through again the
active linked connections.
Fortunately, everytime a new linked connection is created, the event is
activated and thus we would go through the active list again. On a busy relay,
this issue is mitigated by that but on a slower relays or bridge, a connection
could get stuck for a while until a new directory information request would
show up.
Fixes#28717, #28912
The point of this function is to make sure that the ed25519-based
implementation of curve25519_basepoint() actually works when we
start tor, and use the regular fallback implementation if it
doesn't. But it accounts for 9% of our startup time in the case
when we have directory information, and I think it's safe to make
the test shorter. After all, it has yet to find any actual bugs in
curved25519_scalarmult_basepoint_donna() on any platforms.
Closes ticket 28838.
Add the bootstrap tag name to the log messages, so people
troubleshooting connection problems can look up a symbol instead of a
number. Closes ticket 28731.
Merge Phoul's two lists into teor's list.
Replace the 150 fallbacks originally introduced in Tor 0.3.3.1-alpha in
January 2018 (of which ~115 were still functional), with a list of
157 fallbacks (92 new, 65 existing, 85 removed) generated in
December 2018.
Closes ticket 24803.
Replace the 150 fallbacks originally introduced in Tor 0.3.3.1-alpha in
January 2018 (of which ~115 were still functional), with a list of
148 fallbacks (89 new, 59 existing, 91 removed) generated in
December 2018.
Closes ticket 24803.
When retrying all SOCKS connection because new directory information just
arrived, do not BUG() if a connection in state AP_CONN_STATE_RENDDESC_WAIT is
found to have a usable descriptor.
There is a rare case when this can happen as detailed in #28669 so the right
thing to do is put that connection back in circuit wait state so the
descriptor can be retried.
Fixes#28669
Signed-off-by: David Goulet <dgoulet@torproject.org>
This helper function marks an entry connection as pending for a circuit and
changes its state to AP_CONN_STATE_CIRCUIT_WAIT. The timestamps are set to
now() so it can be considered as new.
No behaviour change, this helper function will be used in next commit.
Part of #28669
Signed-off-by: David Goulet <dgoulet@torproject.org>
Use the helper function connection_ap_mark_as_waiting_for_renddesc()
introduced in previous commit everywhere in the code where an AP connection
state is transitionned to AP_CONN_STATE_RENDDESC_WAIT.
Part of #28669
Signed-off-by: David Goulet <dgoulet@torproject.org>
The transition for a connection to either become or go back in
AP_CONN_STATE_RENDDESC_WAIT state must make sure that the entry connection is
_not_ in the waiting for circuit list.
This commit implements the helper function
connection_ap_mark_as_waiting_for_renddesc() that removes the entry connection
from the pending list and then change its state. This code pattern is used in
many places in the code where next commit will remove this code duplication to
use this new helper function.
Part of #28669
Signed-off-by: David Goulet <dgoulet@torproject.org>
To succesful compile tor-print-ed-signing-cert.exe on Windows we
sometimes need to include the @TOR_LIB_GDI@ library.
See: https://bugs.torproject.org/28485
Correctly identify Windows 8.1, Windows 10, and Windows Server 2008
and later from their NT versions.
On recent Windows versions, the GetVersionEx() function may report
an earlier Windows version than the running OS. To avoid user
confusion, add "[or later]" to Tor's version string on affected
versions of Windows.
Remove Windows versions that were never supported by the
GetVersionEx() function.
Stop duplicating the latest Windows version in get_uname().
Fixes bug 28096; bugfix on 0.2.2.34; reported by Keifer Bly.
In conn_close_if_marked(), we can decide to keep a connection open that still
has data to flush on the wire if it is being rate limited on the write side.
However, in this process, we were also looking at the read() side which can
still have token in its bucket and thus not stop the reading. This lead to a
BUG() introduced in 0.3.4.1-alpha that was expecting the read side to be
closed due to the rate limit but which only applies on the write side.
This commit removes any bandwidth check on the read side and simply stop the
read side on the connection regardless of the bucket state. If we keep the
connection open to flush it out before close, we should not read anything.
Fixes#27750
Signed-off-by: David Goulet <dgoulet@torproject.org>
Apparently some freebsd compilers can't tell that 'c' will never
be used uninitialized.
Fixes bug 28413; bugfix on 0.2.9.3-alpha when we added support for
longer AES keys to this function.
We don't use this syscall, but openssl apparently does.
(This syscall puts a socket into a half-closed state. Don't worry:
It doesn't shut down the system or anything.)
Fixes bug 28183; bugfix on 0.2.5.1-alpha where the sandbox was
introduced.
Apparently, even though the manpage says it returns an int, it
can return a long instead and cause a warning.
Bug not in any released Tor. Part of #28399
Resume refusing to start with relative file paths and RunAsDaemon
set (regression from the fix for bug 22731).
Fixes bug 28298; bugfix on 0.3.3.1-alpha.
