FWIW,
It may be that our issue was this:
http://www.freebsd.org/cgi/query-pr.cgi?pr=78776
What I have since then tried:
- setting up irqs as sources for /dev/random pool, (in my case
vmstat -i shows a lot of activity at 2, 10, 0 and 8, and I used the
first two with rndcontrol and then in rc.conf w/ rand_irqs=)
- installing prngd, and using it for SSL seeding
- replacing SSLRandomSeed with different values (egd, builtin,
urandom)
At first. just commenting out the openssl php module didn't work, but
now that I have more activity in /dev/random due to using irqs,
removing openssl php module was apparently a sufficient short-term fix
for my problem at least.. and it works now.
However, I have no idea why it had suddenly ceased working; if I'd
have to guess, this might have had something to with php4's openssl
revision 1.84 at ports/lang/php4/Makefile (a compilation option to
build openssl statically).
HTH..
On Sun, 20 Mar 2005, Pekka Savola wrote:> I've started to experience the same thing which Kyle Mott and
"Vlad" reported
> with apache+mod_ssl crashing when SSL is enabled:
>
> #0 0x2840b63c in engine_table_select () from /usr/lib/libcrypto.so.3
> #1 0x283ebc88 in ENGINE_get_default_RAND () from /usr/lib/libcrypto.so.3
> #2 0x283eaf6e in RAND_get_rand_method () from /usr/lib/libcrypto.so.3
> #3 0x283eb07c in RAND_seed () from /usr/lib/libcrypto.so.3
> #4 0x2830be47 in ssl_rand_seed () from /usr/local/libexec/apache/libssl.so
> #5 0x28307dac in ssl_init_Module () from
/usr/local/libexec/apache/libssl.so
> #6 0x8055714 in ap_init_modules ()
> #7 0x805d8fe in standalone_main ()
> #8 0x805e08b in main ()
> #9 0x804fcde in _start ()
>
> A few salient points:
> - running FreeBSD-4.11 stable, and the latest apache+mod_ssl port w/ PHP
> - the crash occurs before httpd creates the child processes
> - disabling PHP does not work; disabling SSL works around this
> problem (but is not an acceptable solution, of course :)
> - I don't have any cryptocards, nor have I made any significant
> changes to the system since this started happening. The system
> uptime was around 40 days, so the process has at least started fine
> 40 days ago.
> - rebuilding world and apache+mod_ssl does not help
> - /dev/{u,}random seem to be fine
> - my /tmp is mounted nodev,noexec,nosuid but this has worked with it
> in the past, so should not be an issue.
>
>
--
Pekka Savola "You each name yourselves king, yet the
Netcore Oy kingdom bleeds."
Systems. Networks. Security. -- George R.R. Martin: A Clash of Kings