Home
Reading
Searching
Subscribe
Sponsors
Statistics
Posting
Contact
Spam
Lists
Links
About
Hosting
Filtering
Features Download
Marketing
Archives
FAQ
Blog
 
Gmane
From: Matt Mackall <mpm <at> selenic.com>
Subject: Re: [Security] [PATCH] proc: avoid information leaks to non-privileged processes
Newsgroups: gmane.linux.kernel.lsm
Date: Tuesday 5th May 2009 05:50:11 UTC (over 8 years ago)
On Mon, May 04, 2009 at 03:24:15PM -0700, Linus Torvalds wrote:
> 
> 
> On Mon, 4 May 2009, Eric W. Biederman wrote:
> 
> > Arjan van de Ven  writes:
> > 
> > > On Mon, 4 May 2009 12:00:12 -0700 (PDT)
> > > Linus Torvalds  wrote:
> > >
> > >> 
> > >> 
> > >> On Mon, 4 May 2009, Jake Edge wrote:
> > >> >
> > >> > This is essentially v2 of "[PATCH] proc: avoid leaking eip, esp,
or
> > >> > wchan to non-privileged processes", adding some of Eric
Biederman's
> > >> > suggestions as well as the start_stack change (only give out that
> > >> > address if the process is ptrace()-able).  This has been tested
> > >> > with ps and top without any ill effects being seen.
> > >> 
> > >> Looks sane to me. Anybody objects?
> > >> 
> > >
> > > Acked-by: Arjan van de Ven 
> > 
> > Looks sane here.
> > 
> > Acked-by: "Eric W. Biederman" 
> 
> Ok, applied.
> 
> Also, does anybody have any commentary or opinion on the patch by Matt 
> Mackall to use stronger random numbers than "get_random_int()". I wonder 
> what the performance impact of that is - "get_random_int()" is very cheap

> by design, and many users may consider calling "get_random_bytes()" to be

> overkill and a potential performance issue.
> 
> Quite frankly, the way "get_random_bytes()" works now (it does a _full_ 
> sha thing every time), I think it's insane overkill. But I do have to 
> admit that our current "get_random_int()" is insane _underkill_.
> 
> I'd like to improve the latter without going to quie the extreme that 
> matt's patch did.

ASLR aside, get_random_u32 is the right interface (strong,
well-defined type) for random.c to export and get_random_int (weak,
ambiguous type) is the wrong one.

As to what's the appropriate sort of RNG for ASLR to use, finding a
balance between too strong and too weak is tricky. I'd rather move
things to a known safe point and back it off to acceptable performance
from there if anyone complains. So let's do my patch now.

Looking forward:

A faster-but-weakened RNG for ASLR (and similar purposes) will still
need to be strong in the cryptographic sense. Which probably means
having secret state (per cpu?) that changes at every call via a strong
cipher or hash (though lighter than the full RNG). Alternately, we use
a weak primitive (eg halfmd4) with per-task secrets. Not really keen
on the latter as a user may expose outputs across task boundaries that
allow backtracking attacks against the ASLR. In other words, the
latter requires disciplined users, while the former doesn't.

-- 
Mathematics is the supreme nostalgia of our time.
--
To unsubscribe from this list: send the line "unsubscribe
linux-security-module" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
 
CD: 3ms