Re: [DynInst_API:] PIE segfault

[Mohamed Elsabagh <melsabag@xxxxxxx>]

No. Ubuntu 16.04 x86_64.

On Wed, Jun 1, 2016 at 12:47 PM John Detter <jdetter@xxxxxxxx> wrote:

Mohamed,

Is this on Windows?

-- John

On 5/31/2016 6:25 PM, Mohamed Elsabagh wrote:

Hi John,

I pulled the latest commits to master. Now I don't get any errors, but the resulting executable file still doesn't run. It gives the same error: "cannot execute binary file: Exec format error." ÂAny suggestions?

Mohamed

On Tue, May 31, 2016 at 6:24 PM John Detter <jdetter@xxxxxxxx> wrote:

Hey Mohamed,

We found the issue that was causing the assert and we updated the master branch on github. If you want to clone or pull the most recent version it should have the fix for your issue.

Let me know if you have any other issues,

-- John

On 5/31/2016 3:57 PM, Mohamed Elsabagh wrote:

Hi Josh,Â

I just tried again, on a vanilla VM and a fresh clone from https://github.com/dyninst/dyninst.git, and I am still getting the same exact message on stderr:

decodeOneOperand() called with unknown addressing method 18

And the resulting binary does not run: Exec format error.Â

I also tried: `parseThat --binary-edit=ssh.dyn -i 0 /us/bin/ssh`,Âand the resulting binary (ssh.dyn) also fails to load, giving the same Exec format error.

When I run `file ssh.dyn`, the output shows no interpreter.Â

What do you think?

Mohamed

On Tue, May 31, 2016 at 3:26 PM John Detter <jdetter@xxxxxxxx> wrote:

Mohamed,

I did a fresh clone from github and I was able to instrument /usr/bin/ssh on Ubuntu 16.04 server addition, are you sure you are setting your LD_LIBRARY_PATH, C_INCLUDE_PATH and CPLUS_INCLUDE_PATH to the right directories? You may be installing the newly built libraries into a directory that isn't being searched by GCC.

If you are installing globally and you are installing into `/usr/local` you may have to include these directories in your environment:

export LD_LIBRARY_PATH="/usr/local/lib"

export C_INCLUDE_PATH="/usr/local/include"

export CPLUS_INCLUDE_PATH="/usr/local/include"

If you are still having issues, could you send me a tarball of your dyninst directory (source included)? That way I know we're both looking at the same thing.

-- John

On 5/31/2016 12:52 PM, Mohamed Elsabagh wrote:

Yes, that's what I did. I did a fresh clone and install from github.com/dyninst/dyninst. But I am now getting a different error:Â

decodeOneOperand() called with unknown addressing method 18

And even though the output binary is created, it does not execute (exec format error). Â

Again, I am testing with /usr/bin/ssh on Ubuntu 16.04, without any instrumentation (open, get default module, get procedures, save).

Thanks,

Mohamed

On Tue, May 31, 2016 at 1:22 PM John Detter <jdetter@xxxxxxxx> wrote:

Mohamed

git.dyninst.org is now just a mirror and unfortunately it looks like it isn't quite up to date. http://github.com/dyninst/dyninst will get you the latest commits.

If you want to update your origin:

ÂÂÂ ÂÂÂ git remote remove origin

ÂÂÂ ÂÂÂ git remote add origin http://github.com/dyninst/dyninst

ÂÂÂ ÂÂÂ git pull origin master

-- John

On 5/31/2016 11:57 AM, Mohamed Elsabagh wrote:

Hi John,Â

I pulled the latest commit from git.dyninst.org, which resulted in the error in my previous email.Â

Now using a clone from the github path your provided, I am getting the following message on stderr:Â

decodeOneOperand() called with unknown addressing method 18

And even though the output binary is created, it does not execute (exec format error). Â

Again, I am testing with /usr/bin/ssh on Ubuntu 16.04, without any instrumentation (open, get default module, get procedures, save).

Thanks,

Mohamed

On Tue, May 31, 2016 at 11:37 AM John Detter <jdetter@xxxxxxxx> wrote:

Mohamed,

Are you sure you are using the latest master? In my version of
arch-x86.C line 7993 isn't inside the ia32_decode function. Could you
try pulling from master and rebuilding/rerunning? If you could provide
another stack trace that would be really helpful.

-- John

P.S. here is the latest commit information for master
(http://github.com/dyninst/dyninst):

commit df1523dd4003107b959046dd047402642f530c43
Merge: 85cebd3 06c649f
Author: Bill Williams <wwilliam47@xxxxxxxxx>
Date:Â ÂFri May 27 14:37:50 2016 -0500

  ÂMerge pull request #61 from
dyninst/Functions_not_filed_into_correct_Modules

  ÂFix Function/Module mapping

On 5/30/2016 9:06 PM, Mohamed Elsabagh wrote:
> There seems to be a different issue now: calling getProcedures() on
> the default module of a stripped PIE results in an assertion failure
> at common/src/arc-x86.C:7993. It seems that the heuristic gap parser
> is trying to decode the assembly as x86_32 instead of x86_64 (I may be
> wrong though). Exact stack trace is attached.
>
> This is triggered by simply opening the binary, getting the default
> module, then calling getProcedure.
>
> Sample offending program is /usr/bin/ssh on Ubuntu 16.04 x86_64.