Re: libpam segfault when passwd passes NULL pamh (was Re: gcc -O2 produces invalid object code (x86_64, netbsd-5 branch))

[Richard Hansen <rhansen%bbn.com@localhost>]

Joerg Sonnenberger wrote:
> > There's also the issue of whether it's OK to pass random garbage to 
> > pam_strerror().  If it's not OK, then how does one print error 
> > messages for pam_start() failures?  pam_start() doesn't set pamh on 
> > error, so its value could be anything.
>
> Fully agreed. Given that pam_start can only fail due to programming 
> errors or memory allocation, the error message is not very useful 
> anyway. I think the attached patch is the best approach.

There's at least one other way pam_start() can fail:  if you misspell 
the name of a pam module in an /etc/pam.d file.  (Can you guess how I 
originally stumbled across the segfault? :-P)  A "module not found" 
message would be useful.

We could modify pam_start() to set pamh to NULL on error so that 
pam_strerror() does the right thing, or we could do what we're doing now 
and simply ignore pamh in pam_strerror().  See the attached patch.

-Richard