netbsd-5 apache-2.2.14 SIGBUS problem

[Chris Ross <cross+netbsd%distal.com@localhost>]

  This isn't a -current issue, but a netbsd-5 issue, but a recent  
one.  I rebuilt all sets and reinstalled from sources approximately  
January 28, 2010.  This is on a sparc64 box.

  I've built an apache that seems to work just fine most of the  
time.  Basic web works, php5 works, and subversion via https even  
*mostly* works.  However, I have a client that can't establish a  
connection to the host.  I've eventually figured out that the server  
process is dieing via a SIGBUS, thus canceling the attempt to open the  
SSL connection.

  I've finally (after far too long) figured out how to trap the crash  
and get a backtrace.  I see:

Starting program: /usr/pkg/sbin/httpd -X

Program received signal SIGBUS, Bus error.
0x0000000041cae3a4 in _SHA384_Final () from /usr/lib/libc.so.12
(gdb) where
#0  0x0000000041cae3a4 in _SHA384_Final () from /usr/lib/libc.so.12
#1  0x0000000040cec2e8 in EVP_DigestFinal_ex () from /usr/lib/ 
libcrypto.so.4
#2  0x0000000040cde978 in HMAC_Final () from /usr/lib/libcrypto.so.4
#3  0x000000004083f5bc in ssl3_send_newsession_ticket ()
   from /usr/lib/libssl.so.6
#4  0x000000004084088c in ssl3_accept () from /usr/lib/libssl.so.6
#5  0x000000004082a4c0 in SSL_accept () from /usr/lib/libssl.so.6
#6  0x0000000040831b4c in ssl23_get_client_hello () from /usr/lib/ 
libssl.so.6
#7  0x0000000040832370 in ssl23_accept () from /usr/lib/libssl.so.6
#8  0x000000004082a4e4 in SSL_accept () from /usr/lib/libssl.so.6
#9  0x00000000001e20d4 in ssl_io_filter_connect (filter_ctx=0x43578c78)
    at ssl_engine_io.c:1090
#10 0x00000000001e2b5c in ssl_io_filter_input (f=0x43592078,  
bb=0x43595790,
    mode=AP_MODE_GETLINE, block=APR_BLOCK_READ, readbytes=0)
    at ssl_engine_io.c:1336
#11 0x000000000015f9f0 in ap_get_brigade (next=0x43592078,  
bb=0x43595790,
    mode=AP_MODE_GETLINE, block=APR_BLOCK_READ, readbytes=0)
    at util_filter.c:489
#12 0x0000000000135f64 in ap_rgetline_core (s=0x435940d8, n=8192,
    read=0xffffffffffffb1a0, r=0x435940a8, fold=0, bb=0x43595790)
    at protocol.c:231
#13 0x00000000001369fc in read_request_line (r=0x435940a8,  
bb=0x43595790)
    at protocol.c:596
#14 0x000000000013773c in ap_read_request (conn=0x43578298) at  
protocol.c:891
#15 0x00000000001fa320 in ap_process_http_connection (c=0x43578298)
    at http_core.c:183

  So, something about the SSL connection is causing it to crash, but  
apparently it's in libc, rather than one of the openssl libraries.   
Therefore, a general question to netbsd-current seemed appropriate.

  Can anyone guess as to why this might be the case?  Or, suggest  
what it is that this *one* client is doing to cause this problem?

  Any pointers to suggested diagnostic methods also very much welcomed.

                                           - Chris