Bug 1144793 - [abrt] glibc: _dl_check_map_versions(): ld-2.20.so killed by SIGSEGV
Summary: [abrt] glibc: _dl_check_map_versions(): ld-2.20.so killed by SIGSEGV
Keywords:
Status: CLOSED WONTFIX
Alias: None
Product: Fedora
Classification: Fedora
Component: glibc
Version: 21
Hardware: x86_64
OS: Unspecified
unspecified
unspecified
Target Milestone: ---
Assignee: Carlos O'Donell
QA Contact: Fedora Extras Quality Assurance
URL: https://retrace.fedoraproject.org/faf...
Whiteboard: abrt_hash:ef7f1c0ac6c00b415f79b221d29...
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2014-09-21 02:15 UTC by 汪明衡
Modified: 2016-11-24 12:26 UTC (History)
9 users (show)

Fixed In Version:
Clone Of:
Environment:
Last Closed: 2014-10-02 19:12:23 UTC
Type: ---
Embargoed:


Attachments (Terms of Use)
File: backtrace (31.69 KB, text/plain)
2014-09-21 02:15 UTC, 汪明衡
no flags Details
File: cgroup (179 bytes, text/plain)
2014-09-21 02:15 UTC, 汪明衡
no flags Details
File: core_backtrace (2.47 KB, text/plain)
2014-09-21 02:15 UTC, 汪明衡
no flags Details
File: dso_list (1.10 KB, text/plain)
2014-09-21 02:15 UTC, 汪明衡
no flags Details
File: environ (57 bytes, text/plain)
2014-09-21 02:15 UTC, 汪明衡
no flags Details
File: limits (1.29 KB, text/plain)
2014-09-21 02:15 UTC, 汪明衡
no flags Details
File: maps (5.11 KB, text/plain)
2014-09-21 02:15 UTC, 汪明衡
no flags Details
File: open_fds (131 bytes, text/plain)
2014-09-21 02:15 UTC, 汪明衡
no flags Details
File: proc_pid_status (915 bytes, text/plain)
2014-09-21 02:15 UTC, 汪明衡
no flags Details
File: var_log_messages (524 bytes, text/plain)
2014-09-21 02:15 UTC, 汪明衡
no flags Details

Description 汪明衡 2014-09-21 02:15:34 UTC
Version-Release number of selected component:
glibc-2.20-2.fc21

Additional info:
reporter:       libreport-2.2.3
backtrace_rating: 4
cmdline:        /lib64/ld-linux-x86-64.so.2 /usr/bin/glreadtest
crash_function: _dl_check_map_versions
executable:     /usr/lib64/ld-2.20.so
kernel:         3.16.3-300.fc21.x86_64
runlevel:       N 5
type:           CCpp
uid:            0

Truncated backtrace:
Thread no. 1 (8 frames)
 #0 _dl_check_map_versions at dl-version.c:229
 #1 _dl_check_all_versions at dl-version.c:386
 #2 version_check_doit at rtld.c:514
 #3 _dl_receive_error at dl-error.c:217
 #4 dl_main at rtld.c:1671
 #5 _dl_sysdep_start at ../elf/dl-sysdep.c:249
 #6 _dl_start_final at rtld.c:306
 #7 _dl_start at rtld.c:412

Comment 1 汪明衡 2014-09-21 02:15:40 UTC
Created attachment 939668 [details]
File: backtrace

Comment 2 汪明衡 2014-09-21 02:15:41 UTC
Created attachment 939669 [details]
File: cgroup

Comment 3 汪明衡 2014-09-21 02:15:43 UTC
Created attachment 939670 [details]
File: core_backtrace

Comment 4 汪明衡 2014-09-21 02:15:46 UTC
Created attachment 939671 [details]
File: dso_list

Comment 5 汪明衡 2014-09-21 02:15:47 UTC
Created attachment 939672 [details]
File: environ

Comment 6 汪明衡 2014-09-21 02:15:49 UTC
Created attachment 939673 [details]
File: limits

Comment 7 汪明衡 2014-09-21 02:15:51 UTC
Created attachment 939674 [details]
File: maps

Comment 8 汪明衡 2014-09-21 02:15:53 UTC
Created attachment 939675 [details]
File: open_fds

Comment 9 汪明衡 2014-09-21 02:15:55 UTC
Created attachment 939676 [details]
File: proc_pid_status

Comment 10 汪明衡 2014-09-21 02:15:56 UTC
Created attachment 939677 [details]
File: var_log_messages

Comment 11 Carlos O'Donell 2014-10-02 19:12:23 UTC
There are two proprietary libraries loaded from NVIDIA:

0x00007f1b632849b0  0x00007f1b632864b3  Yes (*)     /lib64/tls/libnvidia-tls.so.343.13
0x00007f1b610e8b10  0x00007f1b620046f8  Yes (*)     /lib64/libnvidia-glcore.so.343.13

We have had problems in the past with their libraries, but were never able to track down the problem.

The code in question is attempting to walk all the ELF files and their dependencies and determine if all needed versions are present. There is no reason for this code to fail with a SIGSEGV unless the NVIDIA binaries are corrupt. I would suggest reinstalling them or verifying their md5sum matches the officially distributed binaries from NVIDIA.

I'm going to mark this as CLOSED/WONTFIX. Please reopen if you have a reproducer that someone can run to reproduce the failure.

For example of `ldd /usr/bin/glreadtest` fails, then that's a good reproducer that we can use to inspect what's wrong with the 3rd party DSOs and see if it's a glibc defect.

Comment 12 jan p. springer 2015-02-14 18:35:06 UTC
hi,

i would like to re-open that bug. `ldd /usr/bin/glreadtest` fails, i.e. no output, while `objdump -p /usr/bin/glreadtest` creates output.

i came along this while rebuilding oglplus [http://oglplus.org/]. using system gcc compile/execute tests fail and the binaries behave the same as glreadtest. however, if i configure using clang/clang++ everything is fine. i compared objdump's between gcc and clang versions of the cmake tests and the only difference was that the gcc version required libpthread.so.0 while the clang version did not. also, the clang versions do work, i.e. no SIGSEGV.

for me this started after upgrading to fc21. please indicate what information will be useful for you to start investigating (again).


Note You need to log in before you can comment on or make changes to this bug.