[panda-users] Strange PCs in 64 bit Linux guest

[Igor R]

That's it!
It's quite strange though: I didn't compile the main executable as
position-independent, so I was sure it couldn't be relocated. Besides,
the data segment remained at its default location...
Anyway, disabling ASLR with setarch `uname -m` -R my_prog solved the issue.

2015-06-02 22:20 GMT+03:00 Manolis Stamatogiannakis <mstamat at gmail.com>:
> A guess: Fedora has ASLR enabled. Debian doesn't. GDB turns off ASLR
> (https://outflux.net/blog/archives/2010/07/03/gdb-turns-off-aslr/).
>
> M.
>
> 2015-06-02 10:04 GMT-07:00 Igor R <boost.lists at gmail.com>:
>>
>> Hello,
>>
>> My plugin logs basic-blocks' and some instructions' addresses, which
>> belong to the main executable only. I test it with a simple C program
>> that runs in Debian x86 guest, and the plugin logs the addresses that
>> I'd expect to see.
>> Now I tried to test it with the very same 32-bit program that runs in
>> Fedora x86_64 guest (with glibc.i686 installed), and I was surprised
>> to discover that the instruction addresses that my plugin sees are
>> quite different from the ones I see in gdb (or when reversing
>> statically).
>> I re-compiled the program for 64 bit, but the addresses are still weird.
>> OTOH, if I run the program under gdb, while plugin is active, the
>> addresses are reasonable again.
>>
>> That's said, the program seems to be loaded to the expected base
>> address, because my plugin identifies the target process by
>> identifying a pre-set "cookie" in its data segment - and it *is*
>> there.
>>
>> What could be the reason for such a behavior?
>>
>> Thanks.
>> _______________________________________________
>> panda-users mailing list
>> panda-users at mit.edu
>> http://mailman.mit.edu/mailman/listinfo/panda-users
>
>
>
> _______________________________________________
> panda-users mailing list
> panda-users at mit.edu
> http://mailman.mit.edu/mailman/listinfo/panda-users
>