forbidden memory area access (libCoolProp.so.6)

[niuniuaba]

Hi there,

I am trying out code_saturne 7.2.0 and run into error when define fluid property with CoolProp. By the way, the fluid is just 'Water'.

Code: Select all

command: 
 ./cs_solver --trace --logp --mpi

***************************************************************

                                  (R)
                      Code_Saturne

                      Version 7.2.0


  Copyright (C) 1998-2022 EDF S.A., France

  build Mon 17 Oct 2022 01:44:43 PM CST

  The Code_Saturne CFD tool  is free software;
  you can redistribute it and/or modify it under the terms
  of the GNU General Public License as published by the
  Free Software Foundation; either version 2 of the License,
  or (at your option) any later version.

  The Code_Saturne CFD tool is distributed in the hope that
  it will be useful, but WITHOUT ANY WARRANTY; without even
  the implied warranty of MERCHANTABILITY or FITNESS FOR A
  PARTICULAR PURPOSE.  See the GNU General Public License
  for more details.

***************************************************************


Local case configuration:

  Date:                Mon 17 Oct 2022 01:48:59 PM CST
  System:              Linux 5.15.0-50-generic (Ubuntu 22.04.1 LTS)
  Machine:             qe
  Processor:           model name	: Intel(R) Core(TM) i7-8750H CPU @ 2.20GHz
  Memory:              15652 MB
  User:                wing (wing)
  Directory:           /home/wing/Data/CAE/Pipe/pipeSaturne/Case1/RESU/20221017-1348
  MPI ranks:           4 (appnum attribute: 0)
  OpenMP threads:      2
  Processors/node:     12

  Compilers used for build:
    C compiler:        gcc (Ubuntu 11.2.0-19ubuntu1) 11.2.0
    C++ compiler:      g++ (Ubuntu 11.2.0-19ubuntu1) 11.2.0
    Fortran compiler:  GNU Fortran (Ubuntu 11.2.0-19ubuntu1) 11.2.0

  MPI version: 3.1 (Open MPI 4.1.2)
  OpenMP version: 4.5

  External libraries:
    PT-SCOTCH 6.1.2

  I/O read method:     collective MPI-IO (explicit offsets)
  I/O write method:    collective MPI-IO (explicit offsets)
  I/O rank step:        1


===============================================================

                   CALCULATION PREPARATION
                   =======================


 ===========================================================



SIGSEGV signal (forbidden memory area access) intercepted!

Call stack:
   1: 0x7f5d48b3cfa6 <_ZN8CoolProp16_PropsSI_outputsERSt10shared_ptrINS_13AbstractStat> (libCoolProp.so.6)

I noticed that 'cs_meg_volume_function.c' is not generated. And if I change fluid property method to 'user_material' and set properties manually the calculation executed as expected. It's probably a CoolProp issue but I cannot find an answer on their github site yet and I hope there is some hint here.

Thanks in advance.

[Yvan Fournier]

Hello,

I'll run a test soon to see if the issue is in code_saturne or Coolprop, and keep you updated in a few days at most.

Regards,

Yvan

[niuniuaba]

Hello Yvan,

Thank you for your quick response. Here is some information for reference:
Ubuntu 22.04, gcc 11.2.0, OpenMPI 4.1.2, CoolProp 6.4.1 downloaded from sourceforge.net and built by

Code: Select all

cmake -DCOOLPROP_INSTALL_PREFIX=/home/wing/Apps/CoolProp -DCOOLPROP_SHARED_LIBRARY=ON

Best regards,
niuniuaba

[Yvan Fournier]

Hello,

I can reproduce the error on a simple case, with uniform material properties (meaning code_saturne calls Coolprop only on a size 1 array of values, and then applies the result to all cells).

Checking under a debugger, it seems the arguments passed to CoolProp are OK, as shown in this debugger screenshot:

So this should be reproducible on a simple C++ example, providing the same inputs (useful for a CoolProp bug report).
My build of CoolProp is an older 6.1.1-dev version, and is not a debug build, but this is probably the same issu you observed with 6.4.1...

Perhaps the other back-ends than "HEOS" might provide better results...

So it seems a CoolProp bug-report might be needed...

Regarding 'cs_meg_volume_function.c' not being generated, this is strange. At least on my test build (master branch), they are generated (I did not test with v7.2.0).

Best regards,

Yvan

[niuniuaba]

Hello,

I confirmed that the same build of CoolProp works properly in CS-7.0.0. I prepared debug versions of CS and CoolProp and now get some more information:

Code: Select all

SIGSEGV signal (forbidden memory area access) intercepted!

Call stack:
   1: 0x7fd701a49520 <+0x42520>                       (libc.so.6)
   2: 0x7fd7004b561d <_ZN8CoolProp16_PropsSI_outputsERSt10shared_ptrINS_13AbstractStatYý9'V> (libCoolProp.so.6)
   3: 0x7fd7004b6e4e <_ZN8CoolProp13_PropsSImultiERKSt6vectorINSt7__cxx1112basic_strinYý9'V>> (libCoolProp.so.6)
   4: 0x7fd7004b7432 <_ZN8CoolProp12PropsSImultiERKSt6vectorINSt7__cxx1112basic_stringYý9'V>>> (libCoolProp.so.6)
   5: 0x7fd70213271d <cs_phys_prop_coolprop+0x80e>    (libsaturne-7.2.so)
   6: 0x7fd701f18379 <cs_phys_prop_compute+0x2ca>     (libsaturne-7.2.so)
   7: 0x7fd702b8f8f1 <cs_gui_physical_properties+0x3c4> (libsaturne-7.2.so)
   8: 0x7fd701fdfe89 <iniusi_+0x523>                  (libsaturne-7.2.so)
   9: 0x7fd701fdf38f <initi1_+0x116>                  (libsaturne-7.2.so)
  10: 0x7fd7036627ab <+0x57ab>                        (libcs_solver-7.2.so)
  11: 0x7fd703662e71 <main+0x154>                     (libcs_solver-7.2.so)
  12: 0x7fd701a30d90 <+0x29d90>                       (libc.so.6)
  13: 0x7fd701a30e40 <__libc_start_main+0x80>         (libc.so.6)
  14: 0x562737e2a0a5 <_start+0x25>                    (cs_solver)
End of stack

And if I run cs_solver from inside gdb I get this:

Code: Select all

(gdb) r
Starting program: /home/wing/Data/CAE/Pipe/pipeSaturne/Case1/RESU/7.2/cs_solver 
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
[New Thread 0x7ffff36cf640 (LWP 216351)]
[New Thread 0x7ffff2ece640 (LWP 216352)]
[New Thread 0x7ffff26cd640 (LWP 216353)]
[New Thread 0x7ffff1ecc640 (LWP 216354)]
[New Thread 0x7ffff16cb640 (LWP 216355)]
[New Thread 0x7ffff0eca640 (LWP 216356)]
[New Thread 0x7ffff06c9640 (LWP 216357)]
[New Thread 0x7fffefec8640 (LWP 216358)]
[New Thread 0x7fffef6c7640 (LWP 216359)]
[New Thread 0x7fffeeec6640 (LWP 216360)]
[New Thread 0x7fffee6c5640 (LWP 216361)]

Thread 1 "cs_solver" received signal SIGSEGV, Segmentation fault.
CoolProp::_PropsSI_outputs (State=std::shared_ptr<CoolProp::AbstractState> (use count 1, weak count 0) = {...}, output_parameters=std::vector of length 1, capacity 1 = {...}, input_pair=CoolProp::PT_INPUTS, in1=std::vector of length 0, capacity 0, in2=std::vector of length 0, capacity 0, IO=std::vector of length 1, capacity 1 = {...}) at /home/wing/Downloads/CoolProp.sources/src/CoolProp.cpp:381
381	                    State->update(input_pair, in1[i], in2[i]);

I don't know how to dig deeper because I am not familiar with debugging. Any tips or general description of doing it in gdb will be appreciated.
Could it be the way of calling coolprop from CS? I noticed that cs_coolprop code in 7.3/7.2 is slightly different from that of 7.0.

Best regards,
niuniuaba

[Yvan Fournier]

Hello,

Thanks for the detailed feedback.
I'll check with v7.0. The changes betwen v7.9 and v7.2 were intended to allow the user to set some CoolProp settings, but thereight be a side effect.

I'll check if reverting to the v7.0 code in v7.2 (not using the additional argument) solves the issue or not, and keep you updated.

Best regards,

Yvan

[niuniuaba]

Hello,

I confirmed that regression to code from 7.0 resolves the issue. More specifically, I replaced cs_coolprop.hxx, cs_coolprop.cxx, cs_physical_properties.h and cs_physical_properties.c from 7.0 then 7.2 compiled successfully and the calculation executed as expected (both from command line and gui). Don't know if there is any side effect though.

Best regards