Re: cygheap base mismatch detected

[Charles Wilson <cygwin@cwilson.fastmail.fm>]

On 3/1/2012 7:14 AM, Corinna Vinschen wrote:
> Hmm.  cygcheck loads the Cygwin DLL dynamically.  It does not depend on
> any other Cygwin distro DLL.  But it's started from a Cygwin parent.  So
> the loaded CYgwin DLL checks the layout just like it had been linked
> against.  And apparently it gets rebased at load time.  Which means to
> me, there's another DLL already loaded into the process at an address
> which overlaps with the address space the Cygwin DLL should have been
> loaded to.
> 
> So I guess you just have to find out what is the cause for rebasing
> the Cygwin DLL.  Try VMMap from sysinternals with a cygcheck -svr piped
> into less, so that you can easily observe the process in vmmap.

It appears to be C:\WINDOWS\SysWOW64\PGHook.dll which is part of Avecto
Privilege Guard (which could certainly be classed as a BLODA IMO; thank
you paranoid corporate IT policies...).

For cygcheck, PGHook.dll gets loaded at its desired image base:

$ objdump -p /c/Windows/syswow64/pghook.dll |grep ImageBase
ImageBase               61100000

So, it appears that for normal cygwin processes, PGHook gets rebased
elsewhere, and cygwin "wins."  For instance, /bin/top.exe -> PGHook gets
loaded at 01FD000, b/c cygwin1.dll is already at 0x61000000.  OTOH,
there's nothing in the cygcheck.exe process that prevents PGHook from
getting its desired base addr -- which then conflicts with cygwin's
desired base addr when it is (later) dynloaded.

Is there some workaround that could be used? It's not as if cygcheck is
really trying to initialize and *use* cygwin1.dll facilities, is it? So,
perhaps cygwin1.dll/dcrt0 could ignore the fact that it has been
rebased, when dynamically loaded? (or perhaps, only when dynloaded by
some app named 'cygcheck.exe').

--
Chuck

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple