adamwill / 389-ds-base

Forked from 389-ds-base 4 years ago
Clone

3d04a0e Directory_Server_8_2_Errata_Candidate_15105_20130625

Authored and Committed by mreynolds 10 years ago
    Directory_Server_8_2_Errata_Candidate_15105_20130625
    
    3 fixes: 893178, 954051, and 970934
    
    Bug 893178 - Attribute are not encrypted on a consumer after
     a full initialization
    
    Bug Description:
            During online initialization of a replica encrypted attributes are not encrypted by the import.
            This is because the import job flag job->encrypt is not set.
    
    Fix Description:
            The fix consist to add the config backend attribute "nsslapd-online-import-encrypt" that is by default set to "on".
            During online 'ldbm_back_wire_import' the config attribute is set into the pblock and set into the job->encrypt
    
    https://bugzilla.redhat.com/show_bug.cgi?id=893178
    
    Reviewed by: Rich Meggison (thanks Rich)
    
    Bug 954051 - DS instance crashes under a high load
    
    https://fedorahosted.org/389/ticket/47349
    
    Reviewed by: nkinder (Thanks!)
    
    Branch: Directory_Server_8_2_Branch
    
    Fix Description: handle_new_connection initializes the connection object,
    then calls connection_table_move_connection_on_to_active_list to put it
    on the list of active connections, then unlocks the c_mutex, then calls
    connection_new_private to allocate c_private.  If another thread
    interrupts after the conn has been moved to the active list, but before
    c_private has been allocated, the new conn will be available via
    connection_table_iterate_active_connections where table_iterate_function
    will attempt to dereference the NULL c_private.
    The fix is to move connection_new_private inside the c_mutex lock, and to
    move connection_table_move_connection_on_to_active_list to be the very last
    thing before releasing the c_mutex lock.  Once the conn is on the active
    list it is live and we cannot do anything else to it.
    Note: I have still not been able to reproduce the problem in a non-debug
    optimized build.
    
    Platforms tested: RHEL6 x86_64
    Note: Before patch, server would crash within 5 minutes.  After patch, server
    has been running for several days in customer environment.
    
    Bug 970934 - Schema replication update failed: Invalid
     syntax
    
    https://fedorahosted.org/389/ticket/278
    Resolves: Ticket #278
    
    Bug Description: Schema replication update failed: Invalid syntax
    
    Reviewed by: nkinder (Thanks!)
    
    Branch: Directory_Server_8_2_Branch
    
    Fix Description: Schema replication apparently either sends everything or
    re-reads everything.  unhashed#user#password is an invalid attribute
    name - # is not allowed in an attribute name.  The fix is to explicitly
    allow this particular attribute name.
    
        
file modified
+13 -12
file modified
+5 -0