program hangs in dld

[Dusan Onofer]

we have custom program written in c,
which is linked with oracle 10.2.0 libs.
Program is 32bit, running on
"HP-UX devd3 B.11.11 U 9000/800"
Program should use SHLIB_PATH to locate oracle libraries (located in $ORACLE_HOME/lib32) -> so it's linked with "+s" specified to ld.

When I run program it hangs in:
"bind_symbol" which is in dld -> output is from gdb:
(gdb) info threads
* 4 system thread 3404991 Priority:154 0x76821be8 in __ksleep+0x10 ()
from /usr/lib/libc.2
3 system thread 3404989 Priority:154 0x76821be8 in __ksleep+0x10 ()
from /usr/lib/libc.2
2 system thread 3404990 Priority:179 0xc00207f4 in bind_symbol+0x34 ()
from /usr/lib/dld.sl
1 system thread 3404985 Priority:154 0x7660c720 in ___lwp_wait_sys+0x10 ()
from /usr/lib/libpthread.1
(gdb) thread 2
[Switching to thread 2 (system thread 3404990)]
#0 0xc00207f4 in bind_symbol+0x34 () from /usr/lib/dld.sl
(gdb) bt
#0 0xc00207f4 in bind_symbol+0x34 () from /usr/lib/dld.sl
#1 0xc0025bd0 in bind_module_symbols+0x174 () from /usr/lib/dld.sl
#2 0xc001f45c in dosmartbind+0x80 () from /usr/lib/dld.sl
#3 0xc00251ac in findlibsym+0x70c () from /usr/lib/dld.sl
#4 0xc003e0ec in findexportsym#HLO_CL_#i9_0x0#ib_0x0+0x430 ()
from /usr/lib/dld.sl
#5 0xc00208a8 in bind_symbol+0xe8 () from /usr/lib/dld.sl
#6 0xc0025bd0 in bind_module_symbols+0x174 () from /usr/lib/dld.sl
#7 0xc001f45c in dosmartbind+0x80 () from /usr/lib/dld.sl
#8 0xc00251ac in findlibsym+0x70c () from /usr/lib/dld.sl
#9 0xc003e0ec in findexportsym#HLO_CL_#i9_0x0#ib_0x0+0x430 ()
from /usr/lib/dld.sl
#10 0xc00208a8 in bind_symbol+0xe8 () from /usr/lib/dld.sl
#11 0xc0025bd0 in bind_module_symbols+0x174 () from /usr/lib/dld.sl
#12 0xc001f45c in dosmartbind+0x80 () from /usr/lib/dld.sl
#13 0xc00251ac in findlibsym+0x70c () from /usr/lib/dld.sl
#14 0xc003e0ec in findexportsym#HLO_CL_#i9_0x0#ib_0x0+0x430 ()
from /usr/lib/dld.sl
#15 0xc00208a8 in bind_symbol+0xe8 () from /usr/lib/dld.sl
#16 0xc0025bd0 in bind_module_symbols+0x174 () from /usr/lib/dld.sl
#17 0xc001f45c in dosmartbind+0x80 () from /usr/lib/dld.sl
#18 0xc00251ac in findlibsym+0x70c () from /usr/lib/dld.sl
#19 0xc003e0ec in findexportsym#HLO_CL_#i9_0x0#ib_0x0+0x430 ()
from /usr/lib/dld.sl
.... cut ... this repeats more than 100times ...
#292 0xc001f45c in dosmartbind+0x80 () from /usr/lib/dld.sl
#293 0xc00251ac in findlibsym+0x70c () from /usr/lib/dld.sl
#294 0xc003e0ec in findexportsym#HLO_CL_#i9_0x0#ib_0x0+0x430 ()
from /usr/lib/dld.sl
#295 0xc00208a8 in bind_symbol+0xe8 () from /usr/lib/dld.sl
#296 0xc00260d0 in bind_symbols+0x138 () from /usr/lib/dld.sl
#297 0xc0036b1c in dlopen_override_now+0xb4 () from /usr/lib/dld.sl
#298 0xc0036eec in __dlopen+0x2e4 () from /usr/lib/dld.sl
#299 0x76861e3c in _dlopen+0xcc () from /usr/lib/libdld.2
#300 0x77c33dbc in slpmloclfv+0x60 ()
from /ccbdevd3_oracle10/product/10.2.0/lib32/libclntsh.sl.10.1
#301 0x77c34190 in slpmloc+0x1cc ()


Behavior of program "slightly" changes when I unset SHLIB_PATH. In that case program works fine. I think it's using "last resort" absolute path which is same as specified in SHLIB_PATH to locate oracle's libraries.

It's problem, when I move compiled program to another environment, where "last resort" paths are not valid and SHLIB_PATH doesn't help -- program hangs.

Any ideas what's happening there?

Thanks,

Dusan

[Dennis Handly]

Do you have the latest dld patch installed? PHSS_39077

If this doesn't fix it, you'll have to contact the Response Center.

[Dusan Onofer]

Thanks, but id didn't help. But I've noticed when I link objects with -B immediate program works as expected. I'll contact Response Center.

[ranganath ramachandra · Accepted solution]

can you try increasing the thread stack size ?

[Dusan Onofer]

Bingo!

increasing thread stack size works. I've tryed this also before, but it was not enough. Now I enlarged it even more and it's fine.
Thank you.