Open MPI logo

Open MPI Development Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Development mailing list

Subject: Re: [OMPI devel] 1.7.4rc2r30148 - crash in MPI_Init on Linux/x86
From: Paul Hargrove (phhargrove_at_[hidden])
Date: 2014-01-09 00:27:46


Ralph,

When rebuilding with --enable-debug and the original gcc-4.0.0 the SEGV
returns.
So, the ompi-1.4 in the LD_LIBRARY_PATH was NOT the cause.

Below is a backtrace from gdb which includes line numbers.
The SEGV is in strlen() which suggests a string which lacks
null-termination.

The initial (siginfo) part of the backtrace provided by Open MPI reads:
[pcp-j-6:02741] *** Process received signal ***
[pcp-j-6:02741] Signal: Segmentation fault (11)
[pcp-j-6:02741] Signal code: Address not mapped (1)
[pcp-j-6:02741] Failing at address: 0x63757274

-Paul

#0 0x00a5dbb3 in strlen () from /lib/libc.so.6
#1 0x00a5d8f5 in strdup () from /lib/libc.so.6
#2 0x00534a3b in mca_base_var_enum_create (name=0x349488
"coll_ml_enable_fragmentation_enum",
    values=0x34e014, enumerator=0xbfe03dd0)
    at
/home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/openmpi-1.7.4rc2r30168/opal/mca/base/mca_base_var_enum.c:133
#3 0x0033c328 in mca_coll_ml_register_params ()
    at
/home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/openmpi-1.7.4rc2r30168/ompi/mca/coll/ml/coll_ml_mca.c:257
#4 0x00537585 in register_components (project_name=0x2056f3 "ompi",
type_name=0x2056f8 "coll",
    output_id=-1, src=0xbfe03e7c, dest=0x21bd10)
    at
/home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/openmpi-1.7.4rc2r30168/opal/mca/base/mca_base_components_register.c:116
#5 0x0053736a in mca_base_framework_components_register
(framework=0x21bce0,
    flags=MCA_BASE_REGISTER_DEFAULT)
    at
/home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/openmpi-1.7.4rc2r30168/opal/mca/base/mca_base_components_register.c:67
#6 0x00537ec1 in mca_base_framework_register (framework=0x21bce0,
flags=MCA_BASE_REGISTER_DEFAULT)
    at
/home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/openmpi-1.7.4rc2r30168/opal/mca/base/mca_base_framework.c:107
#7 0x00537f6f in mca_base_framework_open (framework=0x21bce0,
flags=MCA_BASE_OPEN_DEFAULT)
    at
/home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/openmpi-1.7.4rc2r30168/opal/mca/base/mca_base_framework.c:131
#8 0x00152831 in ompi_mpi_init (argc=1, argv=0xbfe04114, requested=0,
provided=0xbfe0400c)
    at
/home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/openmpi-1.7.4rc2r30168/ompi/runtime/ompi_mpi_init.c:555
#9 0x00186ce1 in PMPI_Init (argc=0xbfe04090, argv=0xbfe04094) at pinit.c:84
#10 0x080486e9 in main (argc=1, argv=0xbfe04114) at ring_c.c:19

On Wed, Jan 8, 2014 at 8:45 PM, Paul Hargrove <phhargrove_at_[hidden]> wrote:

> Only takes <30 seconds of typing to start the test and I get email when it
> is done.
> Typing these emails takes more of my time than the actual testing does.
>
> -Paul
>
>
> On Wed, Jan 8, 2014 at 8:35 PM, Ralph Castain <rhc_at_[hidden]> wrote:
>
>> If you have the time, it might be worth nailing it down. However, I'm
>> mindful of all the things you need to do, so please only if you have the
>> time.
>>
>> Thanks
>> Ralph
>>
>> On Jan 8, 2014, at 8:23 PM, Paul Hargrove <phhargrove_at_[hidden]> wrote:
>>
>> Ralph,
>>
>> Building with gcc-4.1.2 fixed the problem for me. I also removed an old
>> install of ompi-1.4 that was in LD_LIBRARY_PATH at build time and might
>> have been a contributing factor. If I'd known earlier that it was there, I
>> wouldn't have reported the problem without first removing it.
>>
>> I can build again with gcc-4.0.0 and --enable-debug if you are still
>> interested in trying to get a line number. This would also determine if
>> LD_LIBRARY_PATH was the true culprit.
>>
>> -Paul [Sent from my phone]
>> On Jan 8, 2014 8:02 PM, "Ralph Castain" <rhc_at_[hidden]> wrote:
>>
>>> Most likely problem is a bad backing store site - any chance you could
>>> give me a line number from this? There are a lot of calls to register
>>> params in that code and I'd need some help in figuring out which one wasn't
>>> right.
>>>
>>>
>>> On Jan 8, 2014, at 6:59 PM, Paul Hargrove <phhargrove_at_[hidden]> wrote:
>>>
>>> I am still testing the current 1.7.4rc tarball on my various systems.
>>> The latest failure (shown below) is a SEGV somewhere below MPI_Init on a
>>> old, but otherwise fairly normal, Linux/x86 (32-bit) system.
>>>
>>> $ /home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/INST/bin/mpirun
>>> -np 1 examples/ring_c
>>> [pcp-j-6:29031] *** Process received signal ***
>>> [pcp-j-6:29031] Signal: Segmentation fault (11)
>>> [pcp-j-6:29031] Signal code: Address not mapped (1)
>>> [pcp-j-6:29031] Failing at address: 0x6c6c6f63
>>> [pcp-j-6:29031] [ 0] [0xbe4440]
>>> [pcp-j-6:29031] [ 1]
>>> /home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/INST/lib/libopen-pal.so.6(mca_base_var_enum_create+0x15d)
>>> [0x2b11ed]
>>> [pcp-j-6:29031] [ 2]
>>> /home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/INST/lib/openmpi/mca_coll_ml.so(mca_coll_ml_register_params+0x639)
>>> [0x440909]
>>> [pcp-j-6:29031] [ 3]
>>> /home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/INST/lib/libopen-pal.so.6(mca_base_framework_components_register+0x14e)
>>> [0x2b2cce]
>>> [pcp-j-6:29031] [ 4]
>>> /home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/INST/lib/libopen-pal.so.6(mca_base_framework_register+0x1b5)
>>> [0x2b32a5]
>>> [pcp-j-6:29031] [ 5]
>>> /home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/INST/lib/libopen-pal.so.6(mca_base_framework_open+0x4e)
>>> [0x2b333e]
>>> [pcp-j-6:29031] [ 6]
>>> /home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/INST/lib/libmpi.so.1(ompi_mpi_init+0x53d)
>>> [0xaf359d]
>>> [pcp-j-6:29031] [ 7]
>>> /home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/INST/lib/libmpi.so.1(MPI_Init+0x13d)
>>> [0xb10d6d]
>>> [pcp-j-6:29031] [ 8] examples/ring_c [0x80486e9]
>>> [pcp-j-6:29031] [ 9] /lib/libc.so.6(__libc_start_main+0xdc) [0x125ebc]
>>> [pcp-j-6:29031] [10] examples/ring_c [0x8048631]
>>> [pcp-j-6:29031] *** End of error message ***
>>>
>>> --------------------------------------------------------------------------
>>> mpirun noticed that process rank 0 with PID 29031 on node pcp-j-6 exited
>>> on signal 11 (Segmentation fault).
>>>
>>> --------------------------------------------------------------------------
>>>
>>> The failure shown is for a singleton run, but np=2 fails as well.
>>>
>>> System info:
>>> $ uname -a
>>> Linux pcp-j-6 2.6.18-238.1.1.el5PAE #1 SMP Tue Jan 18 19:28:42 EST 2011
>>> i686 athlon i386 GNU/Linux
>>> $ gcc --version
>>> gcc (GCC) 4.0.0
>>> Copyright (C) 2005 Free Software Foundation, Inc.
>>> This is free software; see the source for copying conditions. There is
>>> NO
>>> warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR
>>> PURPOSE.
>>>
>>> The only configure argument used was --prefix.
>>>
>>> I was going to attach output from "ompi_info --all", but it SEGV's too!
>>>
>>> $ ompi_info --all
>>> [pcp-j-6:29092] *** Process received signal ***
>>> [pcp-j-6:29092] Signal: Segmentation fault (11)
>>> [pcp-j-6:29092] Signal code: Address not mapped (1)
>>> [pcp-j-6:29092] Failing at address: 0x6c6c6f63
>>> [pcp-j-6:29092] [ 0] [0xd8a440]
>>> [pcp-j-6:29092] [ 1]
>>> /home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/INST/lib/libopen-pal.so.6(mca_base_var_enum_create+0x15d)
>>> [0x2db1ed]
>>> [pcp-j-6:29092] [ 2]
>>> /home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/INST/lib/openmpi/mca_coll_ml.so(mca_coll_ml_register_params+0x639)
>>> [0x48d909]
>>> [pcp-j-6:29092] [ 3]
>>> /home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/INST/lib/libopen-pal.so.6(mca_base_framework_components_register+0x14e)
>>> [0x2dccce]
>>> [pcp-j-6:29092] [ 4]
>>> /home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/INST/lib/libopen-pal.so.6(mca_base_framework_register+0x1b5)
>>> [0x2dd2a5]
>>> [pcp-j-6:29092] [ 5]
>>> /home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/INST/lib/libopen-pal.so.6(opal_info_register_project_frameworks+0x57)
>>> [0x2b83d7]
>>> [pcp-j-6:29092] [ 6]
>>> /home/pcp1/phargrov/OMPI/openmpi-1.7-latest-linux-x86/INST/lib/libmpi.so.1(ompi_info_register_framework_params+0x81)
>>> [0xa69251]
>>> [pcp-j-6:29092] [ 7] ompi_info(main+0x2ba) [0x8049a2a]
>>> [pcp-j-6:29092] [ 8] /lib/libc.so.6(__libc_start_main+0xdc) [0x125ebc]
>>> [pcp-j-6:29092] [ 9] ompi_info [0x80496e1]
>>> [pcp-j-6:29092] *** End of error message ***
>>> Segmentation fault (core dumped)
>>>
>>> I will try again with a newer gcc and report back.
>>>
>>> -Paul
>>>
>>> --
>>> Paul H. Hargrove PHHargrove_at_[hidden]
>>> Future Technologies Group
>>> Computer and Data Sciences Department Tel: +1-510-495-2352
>>> Lawrence Berkeley National Laboratory Fax: +1-510-486-6900
>>> _______________________________________________
>>> devel mailing list
>>> devel_at_[hidden]
>>> http://www.open-mpi.org/mailman/listinfo.cgi/devel
>>>
>>>
>>>
>>> _______________________________________________
>>> devel mailing list
>>> devel_at_[hidden]
>>> http://www.open-mpi.org/mailman/listinfo.cgi/devel
>>>
>> _______________________________________________
>> devel mailing list
>> devel_at_[hidden]
>> http://www.open-mpi.org/mailman/listinfo.cgi/devel
>>
>>
>>
>> _______________________________________________
>> devel mailing list
>> devel_at_[hidden]
>> http://www.open-mpi.org/mailman/listinfo.cgi/devel
>>
>
>
>
> --
> Paul H. Hargrove PHHargrove_at_[hidden]
> Future Technologies Group
> Computer and Data Sciences Department Tel: +1-510-495-2352
> Lawrence Berkeley National Laboratory Fax: +1-510-486-6900
>

-- 
Paul H. Hargrove                          PHHargrove_at_[hidden]
Future Technologies Group
Computer and Data Sciences Department     Tel: +1-510-495-2352
Lawrence Berkeley National Laboratory     Fax: +1-510-486-6900