OpenCSW Bug Tracker


Viewing Issue Advanced Details Jump to Notes ] View Simple ] Issue History ] Print ]
ID Category Severity Reproducibility Date Submitted Last Update
0004607 [libssl_dev] regular use crash sometimes 2010-11-16 19:59 2010-12-18 12:29
Reporter mihailp View Status public  
Assigned To yann
Priority normal Resolution fixed Platform
Status closed   OS
Projection none   OS Version
ETA none Product Build
Summary 0004607: apache + openssl coredumps
Description Server: Solaris 10u3 + latest 10Recommended
Soft: Apache 2.2.15 (compiled by Sun Studio 12 + latest patches)
Soft: openssl 0.9.8n from opencsw


This combination of soft coredumps 20+ times per day under load.
I can't repeat it in test bed.


1.

# mdb core.httpd.11638
Loading modules: [ libc.so.1 libuutil.so.1 ld.so.1 ]
> ::stack
libc.so.1`__lwp_park+0x14(eec75a00, 0, ff033a20, 0, 1c00, 1d3c)
libc.so.1`malloc+0x44(8, 1, d95ac, fe643520, ff0303d8, ff03a5a0)
libcrypto.so.0.9.8`CRYPTO_malloc+0x7c(fe740c04, fe740c04, 19d, 8, fe769608, fed1d948)
libcrypto.so.0.9.8`int_new_ex_data+0x6c(1, 95adf0, 95aed4, 2, 2319f0, fe740c04)
libssl.so.0.9.8`SSL_new+0x220(95adf0, 225e50, 0, 225eb0, fecb4ccc, fecc2a7c)
mod_ssl.so`ssl_init_ssl_connection+0x100(934ae0, 1ea, 6, 0, 934ae0, fed1f89c)
mod_ssl.so`ssl_hook_pre_connection+0x194(934ae0, 934830, 934830, 3ca, 934ad8, 940ed8)
ap_run_pre_connection+0x7c(934ae0, 934830, 1, 1ca, 940ed8, 1)
ap_process_connection+0x24(934ae0, 934830, 934830, 3ca, 934ad8, 940ed8)
process_socket+0xc4(9347f0, 934830, 1, 1ca, 940ed8, 1)
worker_thread+0x2e0(480368, 23e070, 0, 0, 0, 0)
libapr-1.so.0`dummy_worker+0x24(480368, e01fc000, 0, 0, ff2084f8, 1)
libc.so.1`_lwp_start(0, 0, 0, 0, 0, 0)


2.
# mdb core.httpd.5550
Loading modules: [ libc.so.1 libuutil.so.1 ld.so.1 ]
> ::stack
libc.so.1`__lwp_park+0x14(dea8fa00, 0, ff033a20, 0, 1c00, 1d3c)
libc.so.1`free+0x1c(776138, 34, d871c, feca6550, ff0303d8, 67fbc8)
libcrypto.so.0.9.8`CRYPTO_free+0x48(0, 618, 400, fe768ff8, 127b40, 776138)
libssl.so.0.9.8`tls1_setup_key_block+0x128(65bab0, 34, 20, 776138, 14, 67fbc8)
libssl.so.0.9.8`ssl3_accept+0x6e0(65bab0, 21d0, fed0c518, 2000, 21d0, 67fbc8)
libssl.so.0.9.8`ssl23_get_client_hello+0x784(65bab0, b, 2210, 3, fecd6fa0, dcc7b91c)
libssl.so.0.9.8`ssl23_accept+0x29c(65bab0, fed0c518, 0, 225e50, fecd6fa0, 2000)
mod_ssl.so`ssl_io_filter_connect+0x3a0(62fe40, 6a4bf0, 1, 0, 0, 0)
mod_ssl.so`ssl_io_filter_input+0x1a0(6a2de8, 6a4bf0, 1, 0, 0, 0)
ap_get_brigade+0x70(6a2de8, 6a4bf0, 1, 0, 0, 0)
ap_rgetline_core+0x90(6a3e60, 2000, dcc7bc44, 6a3e48, 0, 6a4bf0)
read_request_line+0x7c(6a3e48, 6a4bf0, ffffffff, 0, 0, 6a3ff1)
ap_read_request+0x224(62f918, 3, 0, 1fe, 6336a8, 1)
ap_process_http_connection+0x28(62f918, 62f668, 62f668, 3fe, 62f910, 6336a8)
ap_run_process_connection+0x74(62f918, 62f668, 1, 1fe, 6336a8, 1)
ap_process_connection+0x88(62f918, 62f668, 62f668, 3fe, 62f910, 6336a8)
process_socket+0xc4(62f628, 62f668, 1, 1fe, 6336a8, 1)
worker_thread+0x2e0(4809e8, 23e070, 0, 0, 0, 0)
libapr-1.so.0`dummy_worker+0x24(4809e8, dcc7c000, 0, 0, ff2084f8, 1)
libc.so.1`_lwp_start(0, 0, 0, 0, 0, 0)
>


Is this openssl problem?
I don't have any ideas where start to fix.


Steps To Reproduce
Additional Information Next week I will deploy apache 2.2.17 + openssl 0.9.8o (opencsw) and patch 144488-03
Tags No tags attached.
Attached Files txt file icon kelvin-coredumps-30112010.txt [^] (14,219 bytes) 2010-11-30 11:22

- Relationships

-  Notes
(0008471)
yann (developer)
2010-11-20 13:26

Do you still reproduce this bug with openssl 0.9.8o ?
(0008472)
mihailp (reporter)
2010-11-20 15:22

I will find out it after 29.11
Solaris has latest 10Recommended bundle (6.OCT.2010) with kernel patch 142909-17

p.s. 144488-03 has a very interesting bug fix -
6959312 pthread rwlock race condition issue on multi-CPU configuration
(0008504)
mihailp (reporter)
2010-11-29 08:31

updated:
o) 144488-03
o) 2.2.17
o) latest openssl and expat from opencsw
o) latest glassfish loadbalancer (build 18)

it still coredumps!!!
(0008505)
mihailp (reporter)
2010-11-29 08:32

openssl 0.9.8o
(0008506)
mihailp (reporter)
2010-11-30 11:23

unique back traces of today.
8am-11am - 26 corefiles
(0008532)
mihailp (reporter)
2010-12-07 20:28

no coredumps without *glassfish* loadbalancer for the last 24h.
you could close the report.
(0008587)
yann (developer)
2010-12-18 12:29

Ok, thanks for the update, closing the bug.


Copyright © 2000 - 2008 Mantis Group
Powered by Mantis Bugtracker