Replace the Radiotap parser platform.h file with use of helper functions
from utils/common.h to avoid compiler issues with the updated design and
getting pointers to members of packet structs.
Silence the warning about _next_bitmap assignment. This pointer is
dereferenced only with operations that are safe for unaligned access, so
the compiler warning is not helpful here.
__packed might not be defined in this context, so use STRUCT_PACKED from
utils/common.h.
Fixes: e6ac269433 ("radiotap: Update radiotap parser")
Signed-off-by: Jouni Malinen <j@w1.fi>
It was not easily possible to separate configuration of an interface and
credentials when using the configuration file instead of the control
interface or D-Bus interface for setting up the network profiles. This
makes it hard to distribute configuration across a set of nodes which
use wpa_supplicant without also having to store credentials in the same
file. While this can be solved via scripting, having a native way to
achieve this would be preferable.
Turns out there already is a framework to have external password
storages. It only had a single "test" backend though, which is kind of
an in-memory store which gets initialized with all passwords up front
and is mainly for testing purposes. This isn't really suitable for the
above use case: the backend cannot be initialized as part of the central
configuration given that it needs the credentials, and we want to avoid
scripting.
This commit thus extends the infrastructure to implement a new backend,
which instead uses a simple configuration file containing key-value
pairs. The file follows the format which wpa_supplicant.conf(5) uses:
empty lines and comments are ignored, while passwords can be specified
with simple `password-name=password-value` assignments.
With this new backend, splitting up credentials and configuration
becomes trivial:
# /etc/wpa_supplicant/wpa_supplicant.conf
ext_password_backend=file:/etc/wpa_supplicant/psk.conf
network={
ssid="foobar"
psk=ext:foobar
}
# /etc/wpa_supplicant/psk.conf
foobar=ecdabff9c80632ec6fcffc4a8875e95d45cf93376d3b99da6881298853dc686b
Alternative approaches would be to support including other configuration
files in the main configuration, such that common configuration and
network declarations including credentials are split up into separate
files. But the implementation would probably have been more complex
compared to reusing the already-existing framework for external password
backends.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
The function wpa_config_get_line() is used by the wpa_supplicant config
file parser to retrieve the next non-comment non-blank line. We'll need
the same kind of functionality to implement the file-based external
password backend, so as a preparatory step this commit extracts the
function into its own standalone file in the utils package.
No functional changes are expected from this commit.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
The earlier workaround for UBSAN issues in commit 3b6b3ae581 ("Modify
dl_list_for_each() to not use unaligned access with WPA_TRACE") ended up
using a construction in which the type cast to the containing structure
was compared instead of the struct dl_list pointers. While that worked
around the UBSAN issue, it resulted in a comparison that gcc-10
interprets as being out of bounds for struct dl_list (which it obviously
is since this is to find the start of the containing structure).
Revert that workaround and instead, mark the struct dl_list used within
struct os_alloc_trace to have matching 16 octet alignment as the
containing structure. This is also restoring consistent design for
dl_list_for_each*().
Signed-off-by: Jouni Malinen <j@w1.fi>
Change struct dl_list pointer argument to const in list functions that
do not manipulate the list: dl_list_len() and dl_list_empty().
Signed-off-by: Hai Shalom <haishalom@google.com>
We don't really need to duplicate more of this, so just
move the lib.rules include to the end and do more of the
stuff that's common anyway there.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Derive the library name from the directory name, and let each
library Makefile only declare the objects that are needed.
This reduces duplicate code for the ar call. While at it, also
pretty-print that call.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Now that we no longer leave build artifacts outside the build folder, we
can clean up the gitignore a bit. Also move more things to per-folder
files that we mostly had already anyway.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
This is something I hadn't previously done, but there are
cases where it's needed, e.g., building 'wlantest' and then
one of the tests/fuzzing/*/ projects, they use a different
configuration (fuzzing vs. not fuzzing).
Perhaps more importantly, this gets rid of the last thing
that was dumped into the source directories, apart from
the binaries themselves.
Note that due to the use of thin archives, this required
building with absolute paths.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Instead of building in the source tree, put most object
files into the build/ folder at the root, and put each
thing that's being built into a separate folder.
This then allows us to build hostapd and wpa_supplicant
(or other combinations) without "make clean" inbetween.
For the tests keep the objects in place for now (and to
do that, add the build rule) so that we don't have to
rewrite all of that with $(call BUILDOBJS,...) which is
just noise there.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Currently, wpa_config_set(), the function that sets wpa_supplicant
per-network properties, returns 1 when a property it attempts to set is
unchanged. Its global parallel, wpa_config_process_global(), doesn't do
this even though much of the code is very similar. Change this, and
several of the parser functions, to resemble the per-network parser and
setter functions.
Signed-off-by: Matthew Wang <matthewmwang@chromium.org>
base64_encode_no_lf() is otherwise identical to base64_encode(), but it
does not add line-feeds to split the output.
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
We have practically started requiring some C99 features, so might as
well finally go ahead and bring in the C99 bool as well.
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
This is more consistent with the other eloop registrations and avoids a
theoretical integer overflow with 16-bit int should more than 32767
sockets/signals/events be registered.
Signed-off-by: Jouni Malinen <j@w1.fi>
This is more consistent with the other eloop registrations and avoids a
theoretical integer overflow with 16-bit int should more than 32767
sockets be registered (which is not really going to happen in practice).
Signed-off-by: Jouni Malinen <j@w1.fi>
This avoids a theoretical integer overflow with 16-bit unsigned int
should a certificate be encoded with more that 65535 friendly names or
icons.
Signed-off-by: Jouni Malinen <j@w1.fi>
This is more consistent with the other eloop registrations and avoids a
theoretical integer overflow with 16-bit int (not that there would ever
be more that 32767 signal handlers getting registered).
Signed-off-by: Jouni Malinen <j@w1.fi>
int_array_concat() and int_array_add_unique() could potentially end up
overflowing the int type variable used to calculate their length. While
this is mostly theoretical for platforms that use 32-bit int, there
might be cases where a 16-bit int overflow could be hit. This could
result in accessing memory outside buffer bounds and potentially a
double free when realloc() ends up freeing the buffer.
All current uses of int_array_add_unique() and most uses of
int_array_concat() are currently limited by the buffer limits for the
local configuration parameter or frame length and as such, cannot hit
this overflow cases. The only case where a long enough int_array could
be generated is the combination of scan_freq values for a scan. The
memory and CPU resource needs for generating an int_array with 2^31
entries would not be realistic to hit in practice, but a device using
LP32 data model with 16-bit int could hit this case.
It is better to have more robust checks even if this could not be
reached in practice, so handle cases where more than INT_MAX entries
would be added to an int_array as memory allocation failures instead of
allowing the overflow case to proceed.
Signed-off-by: Jouni Malinen <j@w1.fi>
The resource-load-started cannot be used to replace the older
resource-request-starting signal and as such, the final redirect to the
special http://localhost:12345/ URL did not work. Use the decide-policy
signal for navigation action instead.
Also remove the attempt to modify the request URI from
resource-load-started since that is not going to work either. This is
not really critical for functionality, but could eventually be replaced
with a handler for the WebKitWebPage send-request signal.
Signed-off-by: Jouni Malinen <j@w1.fi>
While the users of os_random() do not really need strong pseudo random
numebrs, there is no significant harm in seeding random() with data from
os_get_random(), i.e., /dev/urandom, to get different sequence of not so
strong pseudo random values from os_random() for each time the process
is started.
Signed-off-by: Jouni Malinen <j@w1.fi>
Use a single block each for webkit and webkit2 signal handlers. This
cleans up browser.c to have clear sections for each webkit API version.
Signed-off-by: Jouni Malinen <j@w1.fi>
Avoid unnecessary warnings from webkit on calling gtk_main_quit() more
than once for a single gtk_main() call. This is also fixing an issue for
a corner case where the very first URL has special purpose (osu:// or
http://localhost:12345).
Signed-off-by: Jouni Malinen <j@w1.fi>
The previous implementation of hovering-over-link signal handler did not
really work with webkit2, so replace this with mouse-target-changed
handler.
Signed-off-by: Jouni Malinen <j@w1.fi>
hs20_web_browser() was previously hardcoded to not perform strict TLS
server validation. Add an argument to this function to allow that
behavior to be configured. The hs20-osu-client users are still using the
old behavior, i.e., not validating server certificates, to be usable for
testing purposes.
Signed-off-by: Jouni Malinen <j@w1.fi>
Some things in bfd.h that we use were renamed, and in the case of
bfd_get_section_vma() a parameter was dropped. Work around this.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Test: wpa_supplicant module tests
../src/utils/utils_module_tests.c:933:7: runtime error: left shift of 1 by 31 places cannot be represented in type 'int'
Signed-off-by: Markus Theil <markus.theil@tu-ilmenau.de>
If hostapd or wpa_supplicant is started with both -s and -f command line
arguments, debug log ended up being written only into syslog and the log
file was left empty. Change this so that the log entries will be written
to both places. Either -s or -f (or both) results in debug log to stdout
being disabled which was already the case.
Signed-off-by: Jouni Malinen <j@w1.fi>
On failure of json_alloc_token(), json_parse() can return without
freeing 'str' previously allocated by json_parse_string(). Fix this
adding proper call to os_free().
Signed-off-by: Davide Caratti <davide.caratti@gmail.com>
Allow any pointer to be used as source for encoding and use char * as
the return value from encoding and input value for decoding to reduce
number of type casts needed in the callers.
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
Allow any pointer to be used as source for encoding and use char * as
the return value from encoding and input value for decoding to reduce
number of type casts needed in the callers.
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
There is no use case for adding padding into the base64url encoded
strings, so remove the unneeded add_pad argument that was hardcoded to 0
in all callers.
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
Cy Schubert
Jouni Malinen
Patrick Steinhardt
Hai Shalom
Johannes Berg
Matthew Wang
Jouni Malinen
Ben Greear
Markus Theil
Davide Caratti