This is not thread safe. If threads are reading from pointers to existing
values they could get a partially updated invalid value.
Refactored assignment to a separate function.
This parser features several improvements over the existing one.
For instance, quoted strings (with escape sequences), unlimited includes,
relaxed newline handling (e.g. at the end of files or before/after { and }),
and the difference between empty and unset values (key = vs. key = "").
It also complains a lot more about invalid syntax. The current one accepts
pretty odd stuff (like settings or sections without name) without any
errors or warnings.
This enumerator is a wrapper around glob(3). If that function is not
supported NULL is returned. If no files match or an error occurs during
the pattern expansion an error is logged and the enumerator simply returns
no items.
RFC: if GLOB_ERR is not supplied glob returns GLOB_NOMATCH if e.g. the
base directory of the pattern does not exist, which would otherwise
result in an error. This way there is at least a clear error message in
case of a typo.
__cmp__() is not supported anymore with Python 3 and cmp() is deprecated.
Instead rich comparisons should be used (only __lt__() is required for
sorting).
This solves a problem with GNAT when compiling charon-tkm as __atomic*
built-ins are only provided in GCC 4.7 and newer.
Currently GNAT 4.6 and GCC 4.7.2 is shipped with Debian wheezy (stable),
as used in the testing environment. So while the configure script correctly
detected the __atomic* built-ins, and defined HAVE_GCC_ATOMIC_OPERATIONS,
this define turned out to be incorrect when charon-tkm was later built
with GNAT.
FreeBSD 10 deprecated the SIOCSIFADDR etc. commands, so we use this
newer command to set the address and netmask. A destination address
is now also required.
Fixes#566.
Adds support for GCC's __atomic* built-ins and improves the performance
of logging (for ignored log levels) and half-open IKE_SA checking under
high loads.
Also fixes two potential race conditions in the load-tester plugin.
For some rwlock_t implementations acquiring the read lock could be quite
expensive even if there are no writers (e.g. because the implementation
requires acquiring a mutex to check for writers) particularly if the
lock is highly contended, like it is for the vlog() method.
This patch is based on one by Christoph Gouault.
Currently, to count the total number of half_open IKE_SAs,
get_half_open_count sums up the count of each segment in the SA hash
table (acquiring a lock for each segment). This procedure does not scale
well when the number of segments increases, as the method is called for
each new negotiation.
Instead, lets maintain a global atomic counter.
This optimization allows the use of big values for charon.ikesa_table_size
and charon.ikesa_table_segments.
These are available since GCC 4.7 and will eventually replace the __sync
operations. They support the memory model defined by C++11. For instance,
by using __ATOMIC_RELAXED for some operations on the reference counters we
can avoid memory barriers, which are required by __sync operations (whose
memory model essentially is __ATOMIC_SEQ_CST).
On many architectures it is safe to read the value directly (those
using cache coherency protocols, and with atomic loads for 32-bit
values) but it is not if that's not the case or if we ever decide to
make refcount_t 64-bit (load not atomic on x86).
So make sure the operation is actually atomic and that users do not
have to care about the size of refcount_t.
Before this change a reqid set on the create_child_t task was used as
indicator of the CHILD_SA being rekeyed. Only if that was not the case
would the local traffic selector be changed to 0.0.0.0/0|::/0 (as we
don't know which virtual IP the gateway will eventually assign).
On the other hand, in case of a rekeying the VIP is expected to remain
the same, so the local TS would simply equal the VIP.
Since c949a4d501 reauthenticated CHILD_SAs also have the reqid
set. Which meant that the local TS would contain the previously
assigned VIP, basically rendering the gateway unable to assign a
different VIP to the client as the resulting TS would not match
the client's proposal anymore.
Fixes#553.
This is undefined behavior as per the C99 standard (sentence 1185):
"If the value of the right operand is negative or is greater or equal
to the width of the promoted left operand, the behavior is undefined."
Apparently shifts may be done modulo the width on some platforms so
a shift by 32 would not shift at all.
The DH transform is optional for ESP/AH proposals. The initiator can
include NONE (0) in its proposal to indicate that while it prefers to
do a DH exchange, the responder may still decide to not do so.
Fixes#532.
But disable the gcrypt plugin, as it causes leaks.
Also disable the backtraces by libunwind as they seem to cause
threads to get cleaned up after the leak detective already has been
disabled, which leads to invalid free()s.
We still try to calculate these directly as it can avoid a dependency on
the pkcs1 or other plugins. But for e.g. PGPv3 keys we need to delegate the
actual fingerprint calculation to the pgp plugin.
Adds a config file and build script for Travis CI. Makes the unit tests
buildable with Clang, and test vectors are now actually verified when
the unit tests are executed.
Also adds options to run only selected test suites and to increase the debug
level during unit tests.
The --enable/disable configure options have been reordered and grouped, and
an option to enable all the features has been added (plus an option to
select a specific printf-hook implementation).
Not sure if 4 jobs is optimal, but according to the docs each build host
has 1.5 virtual cores available (although "getconf _NPROCESSORS_ONLN"
returns 32, which is probably the number of real cores underneath), so
more jobs might not actually reduce the build time much more.
The Blowfish and Twofish implementations provided by the gcrypt plugin
only support specific key lengths, which we don't know when testing
against vectors (either during unit tests or during algorithm
registration). The on_create test with a specific key length will be
skipped anyway, so there is no point in treating this failure differently.
Previously, the {ns}.crypto_test.on_add option had to be enabled to
actually test the algorithms, which we can't enforce for the tests in
the test_runner as the option is already read when the crypto factory
is initialized. Even so, we wouldn't want to do this for every unit
test, which would be the result of enabling that option.
Because GCC does not adhere to the priorities defined for constructors
when building with --enable-monolithic (not sure if it was just luck
that it worked in non-monolithic mode - anyway, it's not very portable)
function registration would fail because the hashtable would not be
created yet.
Clang does not initialize padding in union members so hashing the
complete "arg" union could lead to different hashes if the hashed
plugin_feature_t does not have static storage duration.
Fixes#549.
Older autoconf versions (e.g. on CentOS 6.5) produce an empty else block
for the removed empty argument, which the shell then trips over when
executing ./configure.
Fixes#536.
Tobias Brunner