Re: Broken su in current - trying to fix myself, help needed!

From: 韓家標 Bill Hacker <askbill_at_conducive.net>
Date: Thu, 18 Oct 2007 08:31:58 -0400
Stefan Lambrev wrote:
> Hi Bill,
> 
> 韓家標 Bill Hacker wrote:
>> Stefan Lambrev wrote:
>>> Hi,
>>>
>> *snip*
>>
>>>> I will not be surprised if it occurs when building as an 'ordinary 
>>>> user' and does NOT occur when building as root....
>>>>
>>>> BNL (BSD's Not Linux)....
>>>>
>>
>>> I see something similar on all ports that have OPTIONS (make config).
>>> Here is example (do this as user member of wheel, but not root):
>>>
>>
>> Stop right there. '..NOT root'??
>>
>> Why would I DO that?
> You can do this by mistake for example. When you have 10 terminals 
> sometimes you did not pay enough attention are you root or not

LOL!  trust me to know that one!  50 years since I submitted my first card deck 
to a mainframe, but I did exactly that  - twice, yet - in the last 24 hours..

Including EUID in the 'prompt' just need a hug and kiss, as I use several 
different shells...

> Also you may want only to "read" what is the last configuration of a 
> port using: make config (not configure!)
> and for this you do not have to be root( see permitions of /var/db/ports/)

ACK.

> Also it's a nice feature in FreeBSD ports, so I really do not know why 
> not to use it, as it's a feature, but not a bug.
>

ACK.

> Anyway why or why not does not matter.
> The only think that matter is that doing this trigger the bug in "su".
> Bug that does not exist in 6.2-STABLE or before, and normally bugs are 
> exploited by users that are not root.
> 

What Artem is seeing is not (yet) a 'bug' in su in my mind.

MC is 'in the way' of getting accurate response (smells of the classical DOS 
'pause' when in echo-off, and/or at a point in time when stdio is not connected 
to the VTTY in use).

Unless/until mc is ether sorted or taken out of the loop, the result is not 
conclusive.

IOW - I can reproduce the 'fail-to-complete and say so' easily enough in any CLI 
shell so far mentioned, but I cannot reproduce the 'quietly go away and hide' 
behaviour in a 'raw' shell.

That doesn't mean that su is perfect.

But I'd not waste an su coder's time on su so long as there is a lack of 
transparency / lack of proper error return in mc's script handling.

Separate issue.

> P.S. /usr/ports/Mk/ look for SU_CMD :)
> 

And?

Are you of the opinion that suexec-* et al can over-ride system security when 
invoked by a non-root EUID caller?

I surely hope not...

;-)

Bill
Received on Thu Oct 18 2007 - 10:32:00 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:19 UTC