Nginx with SSL problems on Sparc

[Ondrej Jombik]

We are using nginx on Sparc, it's Debian Linux stable

Linux Sunium 2.6.26-2-sparc64 #1 Mon Jun 21 15:45:39 UTC 2010 sparc64
GNU/Linux

We have Apache2 on port 88 as a single upstream, nginx is serving HTTP
and HTTPS as well (ports 80 & 443).

HTTP is working great, however HTTPS has some problems - every say 5th
request is failed and empty page is served. Nginx logs are full of these
error/warning messages:

2010/07/08 02:08:33 [alert] 479#0: worker process 704 exited on signal 10
*** glibc detected *** nginx: worker process: free(): invalid pointer: 0xfd64cf1c ***
2010/07/08 02:08:38 [alert] 479#0: worker process 705 exited on signal 6
2010/07/08 02:08:39 [alert] 479#0: worker process 706 exited on signal 10
2010/07/08 02:09:02 [alert] 479#0: worker process 714 exited on signal 11
2010/07/08 02:09:03 [alert] 479#0: worker process 732 exited on signal 10
2010/07/08 02:09:14 [alert] 479#0: worker process 736 exited on signal 10
*** glibc detected *** nginx: worker process: corrupted double-linked list: 0x000db1a8 ***
2010/07/08 02:09:44 [alert] 744#0: worker process 745 exited on signal 6
2010/07/08 02:09:57 [alert] 744#0: worker process 749 exited on signal 11
2010/07/08 09:23:35 [crit] 751#0: *104 SSL_write() failed (SSL: error:140D0114:SSL routines:SSL_write:uninitial
ized) while sending to client, client: 10.138.10.136, server: Sunium.Platon.SK, request: "GET --URL-- HTTP/1.1", upstream: "--URL--", host: "db.platon.sk", referrer: "--URL--"
2010/07/08 09:23:35 [crit] 751#0: *104 SSL_shutdown() failed (SSL: error:140E0114:SSL routines:SSL_shutdown:uni
nitialized) while sending to client, client: 10.138.10.136, server: 0.0.0.0:443
2010/07/08 09:23:35 [alert] 744#0: worker process 751 exited on signal 11
2010/07/08 09:23:49 [alert] 744#0: worker process 1658 exited on signal 10
2010/07/08 09:23:49 [alert] 744#0: worker process 1659 exited on signal 11
*** glibc detected *** nginx: worker process: double free or corruption (out): 0x70445ad0 ***
2010/07/08 09:40:27 [alert] 744#0: worker process 1660 exited on signal 6
2010/07/08 09:45:33 [alert] 1907#0: worker process 1908 exited on signal 10
2010/07/08 09:45:34 [alert] 1907#0: worker process 1912 exited on signal 11

There are different signals, different errors, I tried to extract the
most descriptive pattern.

I know this is probably platform dependend, since we are using on many
Intel servers without any problem. But my question is what to do next?
Should I debug, or strace this or just simply send this to Debian
package maintainer?

I tried to upgrade nginx, OpenSSL, libssl, I tried several different
combinations, but with no luck.

Thanks for any help or clue

Ondrej

--
Ondrej JOMBIK
Platon Technologies Ltd., Hlavna 3, Sala SK-92701
+421 903 PLATON - info at platon.org - http://platon.org