[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: [JunkMail] Re: Wishes for set ACLs

Pierangelo Masarati wrote:

Howard Chu wrote:

Argh, it's trickier than that - you also have to protect references to lutil_passwd_sasl_conn. I think we should just make lutil_passwd_sasl_conn a parameter to lutil_passwd, to avoid that issue. Also I wonder if we can just move the passwd mutex itself into lutil_passwd. It seems we need some finer granularity here.

Ain't it protected enough by passwd_mutex? It seems to be used consistently right now (I admit a cleaner API would not require all these external mutexes).

Yes, you're right.

Unfortunately there's no clean way to pass all the context pointers that might possibly be needed in through the existing lutil_passwd API. For the smbk5pwd overlay I hacked this by putting it into thread-specific storage. I think I'll do the same for SLAPD_SPASSWD, to eliminate that issue. And then move the mutex into the chk_crypt routine itself. That will speed up password checking overall, in the default case where crypt / SASL are not being used.

 -- Howard Chu
 Chief Architect, Symas Corp.       Director, Highland Sun
 http://www.symas.com               http://highlandsun.com/hyc
 Symas: Premier OpenSource Development and Support