Re: [Sbcl-devel] sbcl 1.0 hangs when the binary has setuid root set

[Todd S. <ts...@op...>]

Samium Gromoff <_de...@ma...> writes:
> Ingvar wrote:
>> Samium Gromoff wrote:
>> > Juho Snellman wrote:
>> > > Running SBCL with suid sounds like a really bad idea in any case.                                  
>> >                                                                                                      
>> > I am writing a parport-poking piece of code (a JTAG debugger),                                       
>> > which requires extended permissions...                                                               
>> >                                                                                                      
>> > If there is a way of having raw parport access without suid,                                         
>> > i would be happy to know..                                                                           
>>
>> If it presents as a device file it MIGHT be doable to just change permissions
>> on the device (or group or owner). Can't say I've ever poked raw parallel
>> ports under any unix, so wouldn't know off-hand (though I've installed parport
>> printers and they definitely used device files).
>
> Unfortunately, this is supposed to be production software, so requiring
> the end user to make changes to their systems which have non-trivial
> security implications is not acceptable...

That would be funny if it weren't so misguided.  Apparently, you think
that changing the group ownership and permissions on a single device
file has "non-trivial security implications", but installing a
suid-root sbcl doesn't.

You do realize that a suid-root sbcl will allow anyone on the system
to do anything they want as root?  It's effectively the same thing as
installing a setuid root shell.  You can just run SB-EXT:RUN-PROGRAM
on /bin/sh, and you've basically got a root shell.

-- 
Todd Sabin                                          <ts...@op...>