Config file is written to a temp file and then it is renamed to the
original config file. However, it is possible that the rename operation
will be commited to storage while file data will be still in cache
causing original config file to be empty or partially written in case of
a system reboot without a clean shutdown. Make this less likely to occur
by forcing the data to be written to the storage device before renaming
the file.
Signed-off-by: Dmitry Shmidt <dimitrysh@google.com>
64-bit builds with CONFIG_WPA_TRACE=y resulted in the wpabuf pointers
getting misaligned (only 32-bit aligned) and that would result in reads
and writes of unaligned size_t values. Avoid this by indicating explicit
alignment requirement for wpabuf_trace to 8 octets (i.e., there will be
extra four octets of padding in case of 64-bit builds).
Similarly, struct os_alloc_trace resulted in some potential misalignment
cases, e.g., when CONFIG_ACS=y uses a 'long double' variable within
struct hostapd_channel_data. Avoid misalignment issues with explicit
alignment indication.
Signed-off-by: Jouni Malinen <j@w1.fi>
The new TEST_FAIL and GET_FAIL control interface commands can be used
similarly to the earlier TEST_ALLOC_FAIL/GET_ALLOC_FAIL design. The new
version is more generic framework allowing any function to be annotated
for failure testing with the TEST_FAIL() macro. This mechanism is only
available in builds with CONFIG_WPA_TRACE_BFD=y and
CONFIG_TESTING_OPTIONS=y. For other builds, the TEST_FAIL() macro is
defined to return 0 to allow the compiler to remove the test code from
normal production builds.
As the first test site, allow os_get_random() to be marked for failing
based on call backtrace.
Signed-off-by: Jouni Malinen <j@w1.fi>
This makes it somewhat easier to use CONFIG_WPA_TRACE=y build with
external programs that might not be aware of the initialization
requirement, e.g., when linking wpa_ctrl.c with a program that does not
use the os_*() wrappers.
Signed-off-by: Jouni Malinen <j@w1.fi>
The new control interface command TEST_ALLOC_FAIL and GET_ALLOC_FAIL can
now be used to trigger memory allocation failures for testing purposes.
TEST_ALLOC_FAIL sets a failure conditions with
<count>:func[;func][;func]... string and GET_ALLOC_FAIL returns the
current state using the same format. Whenever an allocation is made with
a matching backtrace of calling functions, the count is decremented by
one and once zero is reached, the allocation is forced to fail.
Function names can be prefixed with either '=' or '?' to get different
matching behavior. '=' requires this specific function to be the next
one in the backtrace (i.e., do not skip any other functions in the list
which is the default behavior). '?' allows the function to be optionally
present in the backtrace.
Signed-off-by: Jouni Malinen <j@w1.fi>
This function is meant for comparing passwords or hash values where
difference in execution time could provide external observer information
about the location of the difference in the memory buffers. The return
value does not behave like os_memcmp(), i.e., os_memcmp_const() cannot
be used to sort items into a defined order. Unlike os_memcmp(),
execution time of os_memcmp_const() does not depend on the contents of
the compared memory buffers, but only on the total compared length.
Signed-off-by: Jouni Malinen <j@w1.fi>
Relative time shouldn't be calculated based on gettimeofday
because that clock can jump (e.g., when the time is adjusted
by the system administrator.)
On systems where that is available, use CLOCK_BOOTTIME (on
fairly recent Linux systems, this clock takes into account
the time spend suspended) or CLOCK_MONOTONIC (on Linux and
some POSIX systems, this clock is just freely running with
no adjustments.)
Reported-by: Holger Schurig <holgerschurig@gmail.com>
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
This option can be used only for global parameters that are not going
to be changed from settings.
Signed-off-by: Dmitry Shmidt <dimitrysh@google.com>
Signed-off-by: Iliyan Malchev <malchev@google.com>
ANDROID_SETGROUPS_OVERRIDE macro can now be used to override setgroups()
values based on build configuration.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
trace.h may end up including system header files that define offsetof,
so include the compatibility definition from list.h only after this.
Signed-hostap: Jouni Malinen <j@w1.fi>
os_*() wrappers should be used instead of functions from time.h.
Removing the header from includes.h enforces this. os_unix.c can
include this its uses are valid wrapper calls. wps_upnp.c uses
gmtime() for which there is no os_*() wrapper available yet, so
allow it to use time.h, too. Similarly, allow dump_state.c to
use time.h for ctime().
In theory, these calls could fail, but it is not really likely to
happen in practice in the use case here. Anyway, check that they do
not return an error before accepting the length of the file.
This patch adds support for wired IEEE 802.1X client on the Solaris.
I have tested with these:
OS : OpenSolaris 2009.06
EAP : EAP-MD5
Switch : Cisco Catalyst 2950
The arithmetic on void pointer exists in trace routine. On GNU C, it
works because void pointer size is 1, but not all compilers behave like
this. So this patch specifies the size of the pointer.
This WPA_TRACE=y additions allows components to register active references
to memory that has been provided to them as a pointer. If such an actively
referenced memory area is freed, tracer will report this as an error and
backtraces of both the invalid free and the location where this pointer
was marked referenced are shown.
WPA_TRACE=y builds will now verify that memory allocation in done
consistently using os_{zalloc,malloc,realloc,strdup,free} (i.e., no
mixing of os_* functions and unwrapper functions). In addition, some
common memory allocation issues (double-free, memory leaks, etc.) are
detected automatically.
This is a patch for OpenBSD wired IEEE 802.1X. This is only for wired,
not wireless, because OpenBSD uses wpa_supplicant only on wired now.
http://www.openbsd.org/cgi-bin/cvsweb/ports/security/wpa_supplicant/
I have tested with these.
OS : OpenBSD 4.5
EAP : EAP-TLS
Switch : CentreCOM 8724SL
