Re: How trustworthy is that I/O device?

I'd like to have an option like that - it would be nice in instances where
I know a file system is full of errors but want to recover what I can
without a panic, and on the other side of the coin want to panic before
tolerating any detectable inconsistency. This probably digresses a bit from
the security aspects, but I think that sort of option would be of great
general benefit as well as being useful from a security perspective.

What I would personally love to see, from a security perspective, is the
ability to define what sorts of devices I do trust. Does NetBSD have a
mechanism similar to the /dev/disk/by-uuid/ presented in Linux? Taking it a
bit further.. wouldn't it be possible to, for instance, define a certain
vendor:device combo as "accepted" for a USB serial device, and have that be
to the exclusion of others?
I don't wish to spend NetBSD dev time on my own convenience and laziness,
and would happily help out - these simply spring to mind as things I have,
in Linux, taken shortcuts with in the name of convenience that might
(might!) offer some negligible protection (for example, only a uplcom(4)
serial device would be able to fool one of my kernels... that sort of
thing.)

I agree with Matt (both of them) and Erik, and in the spirit of waxing
philosophic, will say that I find it very interesting to find a balance
here given that as a security professional I am also guilty of taking the
easy way out when there is one.



On Wed, Nov 6, 2013 at 1:31 PM, matthew green <[email protected]> wrote:

>
> > What I am saying is that you shouldn't trust it.  We shouldn't have
> > underlying assumptions in the code.  We don't in the networking
> > code.  Treat it as if it's probably trying to cause a denial of
> > service.  I'm saying don't panic, either forcibly unmount or treat
> > it as uncacheable and read-only.
>
> solaris introduced "onerror" mount option, that takes options
> like "panic", "unmount", "readonly", etc.
>
> i've always thought that was a good idea we should copy.
>
>
> .mrg.
>



-- 
Sean