John:

Thanks for explaining the security ramifications of this patch.  I
understand better now, and it sounds like we are doing the right thing.

>> 3) Even though this is TSOL related code, I don't think using "tsol" in
>>     the function name is a good idea.  This is a general patch to tighten
>>     the way modules are loaded.
> 
> Not just an issue within this patch.  We are attempting to manage the
> name space for the TX changes.  This is a requirement by the committee.

It still seems to me that we could probably rework this patch so it is
more a more generic "tightening" of security that we could get upstream
into the glib code if we wanted to.  Perhaps this would require making
it a bit more configurable, so it could work in different ways based
on a configuration choice.  But if we want to maintain our own
TSOL patches for this, I understand the TSOL unique requirements might
make this easier to manage our own patches.

Brian


>>> I thought I'd start you off with one of our more simple Trusted
>>> Extensions patches. :)
>>>
>>> Please take a look at the following NEW patch for glib.
>>>
>>> This patch is to improve the security of the module loading aspect of
>>> glib so that trusted code can not be extended by arbitrary modules and
>>> left open to the case where rogue code can be inserted into a trusted
>>> component.
>>>
>>> In a trusted multi-label session, indicated by the env var
>>> TRUSTED_SESSION, glib will restrict the locations that it will accept
>>> modules from by first asking the runtime linker for the information it
>>> has been configured with and falling back to some default file system
>>> locations.
>>>
>>> Thanks,
>>>
>>> Stephen.
>>
>>
> 


Reply via email to