[krbdev.mit.edu #1214] AutoReply: create_history_entry() attempts to use NULL key

[darrenr@chiron.nabaus.com.au via RT]

This bug has its roots in the UMICH replication code.  The following patch
was used against svr_principal.c to prevent the core dumps.  This should
probably be sent back to the UMICH folks for review (this is just a work-
around without any investigation about correctness) now that the real
culprit has been identified.

Index: svr_principal.c
===================================================================
RCS file: /disk3/CVS/Krb5/src/lib/kadm5/srv/svr_principal.c,v
retrieving revision 1.2
diff -c -r1.2 svr_principal.c
*** svr_principal.c	2002/05/31 00:12:16	1.2
--- svr_principal.c	2002/10/09 07:25:16
***************
*** 939,944 ****
--- 939,946 ----
      krb5_error_code ret;
  
      for (x = 0; x < n_new_key_data; x++) {
+ 	 if (new_key_data[x].key_data_length[0] == 0)
+ 	    continue;
  	 if (ret = krb5_dbekd_decrypt_key_data(context,
  					       &master_keyblock,
  					       &(new_key_data[x]),
***************
*** 946,951 ****
--- 948,955 ----
  	    return(ret);
  	for (y = 0; y < n_pw_hist_data; y++) {
  	     for (z = 0; z < pw_hist_data[y].n_key_data; z++) {
+ 		  if (pw_hist_data[y].key_data[z].key_data_length[0] == 0)
+ 		      continue;
  		  if (ret =
  		      krb5_dbekd_decrypt_key_data(context,
  						  hist_keyblock,
***************
*** 1004,1009 ****
--- 1008,1015 ----
       memset(hist->key_data, 0, n_key_data*sizeof(krb5_key_data));
  
       for (i = 0; i < n_key_data; i++) {
+ 	  if (key_data[i].key_data_length[0] == 0)
+ 	       continue;
  	  if (ret = krb5_dbekd_decrypt_key_data(context,
  						&master_keyblock,
  						&key_data[i],