Potential reasons dlopen could segfault?

Question

What are some of the reasons dlopen could segfault besides the shared object not existing?

In my case, I know the shared object exists, but when my program goes to load it using dlopen, it segfaults. I checked in my lib folder and the shared object is there and the paths are all correct.

    handle = dlopen(libraryName.c_str(), RTLD_LAZY | RTLD_GLOBAL);

gdb bt:

#0  0x00000000001b94f5 in ?? ()
#1  0x00007fffefd96db6 in __do_global_ctors_aux () from /usr/local/lib/MY_LIB2.so
#2  0x00007fffefcf82c3 in _init () from /usr/local/lib/MY_LIB2.so
#3  0x00007fffed69c6c8 in ?? () from /usr/local/lib/MY_LIB1.so
#4  0x00007ffff7de9dc4 in call_init () from /lib64/ld-linux-x86-64.so.2
#5  0x00007ffff7de9ef6 in _dl_init_internal () from /lib64/ld-linux-x86-64.so.2
#6  0x00007ffff7dedf43 in dl_open_worker () from /lib64/ld-linux-x86-64.so.2
#7  0x00007ffff7de9c36 in _dl_catch_error () from /lib64/ld-linux-x86-64.so.2
#8  0x00007ffff7ded7ca in _dl_open () from /lib64/ld-linux-x86-64.so.2
#9  0x00007ffff5c5af26 in dlopen_doit () from /lib64/libdl.so.2
#10 0x00007ffff7de9c36 in _dl_catch_error () from /lib64/ld-linux-x86-64.so.2
#11 0x00007ffff5c5b4cf in _dlerror_run () from /lib64/libdl.so.2 
#12 0x00007ffff5c5afc1 in dlopen@@GLIBC_2.2.5 () from /lib64/libdl.so.2
#13 0x00007ffff6ecef7e in mynamespace::Factory::attachModule (this=0x61d440,    libraryName=...) at Factory.cpp:324
#14 0x00007ffff6ecefe6 in mynamespace::Factory::attachFunction (this=0x61d440, functionName=..., moduleName=...) at Factory.cpp:343
#15 0x00007ffff6ecdd16 in mynamespace::Factory::ReadFile (this=0x61d440, x=...) at Factory.cpp:111
#16 0x00007ffff6ecda62 in mynamespace::Factory::ReadDirectory (this=0x61d440, x=...) at Factory.cpp:79
#17 0x00007ffff6ecdc66 in mynamespace::Factory::ReadDirectory (this=0x61d440, x=0x417901 "/usr/local/lib/") at Factory.cpp:105

#18 0x0000000000410637 in main (argc=2, argv=0x7fffffffdd58) at main.cpp:78

`Factory.cpp:324`, `Factory.cpp:343`, `Factory.cpp:111`, `Factory.cpp:79`, `Factory.cpp:105`. Have you checked your code there already? Have you looked at the [manpage](http://linux.die.net/man/3/dlopen) already? Unfortunately, most of us have not mastered clairvoyance yet. — Sebastian Mach, Aug 13 '12 at 13:59
Yes. library name and all passes through fine. And I have looked at dlopen manpage. Doesn't seem to be much that could help me. Are there any usage requirements that is needed for dlopen to work? — user1496542, Aug 13 '12 at 14:00
We have two possibilities: Close this question. Or you post some of the relevant code. Because possible reasons for segfaulting include bad programming, bugfull libraries, corrupt hardware and solar winds. — Sebastian Mach, Aug 13 '12 at 14:02

score 6 · Accepted Answer · answered Aug 13 '12 at 14:21

In addition to loading a library into memory and fixing up references, the runtime linker runs initializers such as functions labelled __attribute__((constructor)), an init function (if specified with -Wl,-init,...), and constructors for global objects. Your backtrace indicates that one of those is failing.

Specifically, __do_global_ctors_aux runs constructors for global objects. Check those.

Potential reasons dlopen could segfault?

1 Answers1