* Make sure POSIX has the new extension command enumeration function.
* Add support for deinit of extensions.
* Make sure extensions are tracked like they in Windows.
* Fix up a few export definitions.
* Stop using strncpy_s in POSIX code.
* bury common _snprintf that snuck in while testing the Windows version back
into the platform-specific code.
* remove now-unneeded separator defines
* don't free a stack variable on windows
Separate the stdapi handling code from the OS-dependent code. This makes
testing and maintaining the code easier.
This also happens to fix a number of bugs as a side-effect, because it is
clearer what is happening now.
This commit contains a bunch of code tidying (formatting, spaces, naming, etc) as well as new exports for each of the modules so that the extension can be identified. The plan is for the loader to know which modules are loaded so that when stageless meterpreter fires up MSF can query the existing extensions and load the appropriate functionality on the client side.
By making this a static _inline, it is not necessary to guard it, since
an inline is only instantiated if it is used. This also allows adding
one-off debug message for use during debugging sessions, without turning
on DEBUGTRACE all over the place.
Convert a few of the extensions to also do this as well, making them perhaps
slightly smaller.
I am curious why Windows builds define debug this way, vs posix that
just includes it in common.c. Could I just do that instead, assuming
there's no historical reason.
Finally, correct the docs in the posix version of real_dprintf.
If a user attempts to start the clipboard monitor when it is already started then the code path that is taken results in the current clipboard monitor state pointers being lost. The net effect of this is that the existing monitor thread will never be shut down. Not a good thing!
This code fixes that case so that the monitor doesn't create a new monitor thread and doesn't reset important pointers to NULL.
This change also results in a "success" status being returned to the caller. This means it looks like the clipboard monitor has been started even if it was already running. I think this is acceptable and is better than an obscure error.
- try to share some bits between different makefiles, make modifying
global compiler flags not such a huge pain.
- directly specify we should be using the gold rather than bpf linker
- make compiler output largely quiet except where we care - allow
warnings to actually be visible
- don't delete downloaded tarballs with --really-clean
- add missing dependencies between libraries
(--no-add-needed/--no-copy-dt-needed-entries causes lots of trouble)
- update readme to show what to install to build
I made minimal changes to the loader makefile - it breaks easily.
-Os prevents if from being able to load libc, for instance
This adds registry access methods that do an atomic open/<action>/close on
registry keys. They improve efficiency and safety, since we're not passing
HKEY's back and forth to enumerate or read registry keys. This fits the common
use pattern in MSF better anyway.
Now includes group nesting which will require work on the MSF side to
fully support it. But this option is way better as we don't lose info
thanks to serialising to strings.
Added support for a the getsid command so that it's easy to
determine the SID of the current user. This will be useful for
a number of different use cases.
This comes with some other changes as well which include code changes that
were probably made to the kerberos code prior to moving to github. As a result,
the actual changset was lots because the history of mimikatz wasn't imported
into github, and the change history is no longer available on google code.
The goal is to avoid pointer truncation where possible so this commit
changes parameter types to qword where it makes the most sense. This
includes all handles (event, process, thread, registry), addresses
and generic parameters.
