How many users can be created on an AIX V4.3.2 ?

How many users can be created on an AIX V4.3.2 ?

Post by Christophe BRIDE » Thu, 30 Mar 2000 04:00:00



I've just have a problem when I create a new user : I have serevals
thousands users in one group, and I can't create any new user in this group.
This don't worried me but I'm afraid if they are a limitation in the users
number (I must create about 10000 users on my AIX).

Thanks
Christophe BRIDET

 
 
 

How many users can be created on an AIX V4.3.2 ?

Post by Bruce Spence » Sat, 01 Apr 2000 04:00:00


See "mkpasswd" for creating an indexed password database


>I've just have a problem when I create a new user : I have serevals
>thousands users in one group, and I can't create any new user in this
group.
>This don't worried me but I'm afraid if they are a limitation in the users
>number (I must create about 10000 users on my AIX).

>Thanks
>Christophe BRIDET


 
 
 

How many users can be created on an AIX V4.3.2 ?

Post by Norman Levi » Thu, 06 Apr 2000 04:00:00


But if you allow a user to be added to the system and to a primary group and you do not
update /etc/group, haven't you lost one bit of validation that the user was really supposed
to be in that group?  For example, you walk away from your terminal as root, I edit /etc/passwd
and change my group id of 1(staff) to 0(system) and I and not listed in /etc/group.

Seems like a problem?




> > > X-No-Archive: YES
> > > Christophe is describing an entirely different problem --
> > > when you put too many users into a single group the group
> > > entry gets rather large and then the system breaks because
> > > the line is too long.  The hashed password files (AIX doesn't
> > > have hashed group files -- perhaps you are thinking of
> > > Linux) can help if the /etc/passwd file is huge, but that's
> > > it.

> > > There is an APAR coming out Real Soon Now which deals
> > > with the problem of mkuser adding users to their primary
> > > group by default.  As always, I don't have the number, but if
> > > you pester me I might be able to dig it up.

> > Wow.  That problem's still around?  Why's it taken so long to fix
> > it and what are y'all gonna do to fix it?  It's been around since at
> > least the early days of 4.3.x.

> You have to understand that it isn't a "problem", it's one of those
> anomolous behaviours that people didn't consider 11 years ago when
> AIX v3 was being designed.  As such, it isn't really a "bug" -- the
> mkuser command was working exactly the way we wanted it to work
> and we had very good reasons.

> What happened is that system administrators, instead of creating
> groups to subdivide their users went off and put them all into one
> group (which is an administrative mistake -- it becomes the
> "everyone" group and that doesn't help with access control policies).

> Several releases back we changed mkuser due to customer requests
> to not add the user's name to /etc/group for their primary group.
> Well, some customer's complained about this change and it was
> changed back.

> When this issue again arose a decision was made to do something
> that would potentially make both sides happy -- add an option which
> allows the administrator to choose the behavior they want.

> SO .... the APAR, whose number is IY05836, adds a "configuration
> option" which lets the administrator decide -- add the user to their
> primary group in /etc/group or don't.  The upside is that you get
> whichever behavior =you= want.

> (Of course, there =is= a m*to this story -- we will (generally
> speaking) change things if we know what you want)

> --
> Julianne Frances Haugh                 "Life is either a daring adventure, or
> RS/6000 Security Development                 nothing at all."
> AIX Security Development                                -- Helen Keller
> http://www.veryComputer.com/

--
Norman Levin
 
 
 

How many users can be created on an AIX V4.3.2 ?

Post by Norman Levi » Mon, 10 Apr 2000 04:00:00



> X-No-Archive: YES



> > But if you allow a user to be added to the system and to a primary group
> and you do not
> > update /etc/group, haven't you lost one bit of validation that the user
> was really supposed
> > to be in that group?  For example, you walk away from your terminal as
> root, I edit /etc/passwd
> > and change my group id of 1(staff) to 0(system) and I and not listed in
> /etc/group.

> And if I instead write a shell script which changes both at
> once with a single command, haven't I defeated having
> the user in /etc/group as well as /etc/passwd?

** that is true, but the more I get into security in unix, the more I think
it is an oxymoron and anything you can do that can cross reference something
else MIGHT give you a bit more security.  Maybe the only reason to get into
unix security is, the group hold their meetings in the BEST locations. <grin>

Quote:> Some things can't be fixed with software.  They can
> only be fixed by chaining sysadmins to their chairs ;-)

> -- Julie.

--
Norman Levin
 
 
 

How many users can be created on an AIX V4.3.2 ?

Post by Norman Levi » Mon, 17 Apr 2000 04:00:00



> X-No-Archive: YES



> > > And if I instead write a shell script which changes both at
> > > once with a single command, haven't I defeated having
> > > the user in /etc/group as well as /etc/passwd?
> > ** that is true, but the more I get into security in unix, the more I think
> > it is an oxymoron and anything you can do that can cross reference something
> > else MIGHT give you a bit more security.  Maybe the only reason to get into
> > unix security is, the group hold their meetings in the BEST locations. <grin>

> You're starting to get into this whole fuzzy domain of "security
> by obscurity".  "If I can just make it a bit harder, maybe then ..."

** security by obsurity is more like not documenting something (see prior bootinfo
discussion).  This is perhaps 'over documenting?'  I'm suggesting basic cross checks
aren't bad - but certainly not fool proof (and I am the fool to prove it).

- Show quoted text -

Quote:

> Once I know the rules ("edit /etc/passwd then /etc/group") I can
> find a new attack.  And it's really pretty simple --

> #!/bin/ksh
> ed - /etc/passwd << EOF
> /^jfh:/s/^\([^:]*\):\([^:]*\):\([^:]*\):\([^:]*\):/\1:\2:\3:0:/
> w
> q
> EOF
> ed - /etc/group << EOF
> /^system:/s/$/,jfh/
> w
> q
> EOF

> I'm not sure that actually =works=, but it was pretty simple and
> shouldn't be too far from the mark.  And if I can that little shell
> script some place where I can trick you into running it (hmmmm, what
> if I name it "doom" ...),

** I like the prior example, but '*' or 'sleaze' would probably appeal
to me more than 'doom' - woops - many of my students might read this...

it really doesn't matter.

Quote:

> The best rule really is chain the sysadmin to her desk and don't
> let her leave it unless she logs out first.

> What you want really is an entirely different approach.  Things
> like "certain commands can only be run by root from the console"

** now you are chaining things down to hardware and I can't believe that
a well protected system (OS/390) can't do better than insist that certain
commands work only from certain terminals/addresses.

Quote:> and "certain files can only be modified in single user mode from
> the console."  This starts to put the burden on things other than
> obfuscation -- like the badge reader on the machine room door along
> with those really nice video surveilence cameras pointing at whoever
> walks in.

> So ... if'n yer still interested, ya gotta resume you wanna send?
> Hmmmmmmm?!?

** after 27 years with IBM, I kind of enjoy being a consultant,
that is 'being between jobs' in IBM speak <grin>.
--
Norman Levin