From owner-freebsd-hackers  Sun Mar  3 19:43:21 1996
Return-Path: owner-hackers
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.3/8.7.3) id TAA29692
          for hackers-outgoing; Sun, 3 Mar 1996 19:43:21 -0800 (PST)
Received: from ref.tfs.com (ref.tfs.com [140.145.254.251])
          by freefall.freebsd.org (8.7.3/8.7.3) with ESMTP id TAA29679
          for <hackers@freebsd.org>; Sun, 3 Mar 1996 19:43:17 -0800 (PST)
Received: (from julian@localhost) by ref.tfs.com (8.7.3/8.6.9) id TAA27629; Sun, 3 Mar 1996 19:41:43 -0800 (PST)
Message-Id: <199603040341.TAA27629@ref.tfs.com>
Subject: Re: pwd_mkdb - really slow :(
To: rashid@rk.ios.com (Rashid  Karimov)
Date: Sun, 3 Mar 1996 19:41:43 -0800 (PST)
From: "JULIAN Elischer" <julian@ref.tfs.com>
Cc: hackers@freebsd.org
In-Reply-To: <199603040253.VAA07580@rk.ios.com> from "Rashid  Karimov" at Mar 3, 96 09:53:17 pm
X-Mailer: ELM [version 2.4 PL25 ME8b]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: owner-hackers@freebsd.org
Precedence: bulk

> 
> 		HI there ppl,
> 
> 
As I understand it, changes were made to do single-point
changes, and to speed all this up in the 'passwd' case.

to speed up mkpwd you would need a 'diff' file of 
before and after' and do the changes as a set of single-point changes.
This requires that you trust the passwd.master and the database
files to be correctly in sync.


> 
> 
> 	I have quite afew systems here with ~10.00 accounts
> 	on ea ... and pwd_mkdb is _very slow now ... Takes
> 	2+ minutes to rebuild the damn databases every time.
> 
> 	I'm not concerned with passwd command , even though
> 	as I understand it doesn;t do one-point update so far
> 	and rebuilds everything from a scratch, but I'm trying
> 	to somehow speed up the pwd_mkdb itself.
> 
> 	So they're a few ways I can see looking at the source code . 
> 	For example it does two passes on a master password file
> 	it's fed as a parameter - once for unsecure and once
> 	for secure databases - so this could be made only once
> 	and we can write in to db's in parallel, Secondly may
> 	be I'll be able to reuse at least parts of the structure
> 	used for db->put for secure and usecure db's.
> 	And aslo - does any of programs actually use pwd info
> 	indexed by line number ? Do we really need this pass 
> 	( there're actually two passes) ? 
> 
> 	I want to check the the thing using profiler , just 
> 	to see what eats most time in the program ( I do have
> 	a feeling already that it's all those dp->put's ).
these might of course speed things up..

it might also be quicker to build the entire databse in ram 
and bang it out to disk as a single write.. :)

> 
> 
> 	Rashid.
>