Maybe it is silly to ask but,
is there a default protection for buffer overflows in kernels versioned above 2.4.20 ??
sincerely.
On Mon, 2003-05-19 at 00:27, Halil Demirezen wrote:
> Maybe it is silly to ask but,
> is there a default protection for buffer overflows in kernels versioned above 2.4.20 ??
I think this link will be of your interest:
http://linuxtoday.com/security/2003050501026OSKNDV
On Mon, 19 May 2003 01:27:42 +0300
Halil Demirezen <[email protected]> wrote:
> Maybe it is silly to ask but,
> is there a default protection for buffer overflows in kernels versioned above 2.4.20 ??
http://marc.theaimsgroup.com/?l=linux-kernel&m=105189350008143&w=2
is quite interesting
Kudos to the redhat people :)
> http://marc.theaimsgroup.com/?l=linux-kernel&m=105189350008143&w=2
> is quite interesting
yes that is interesting, however, what i want to learn, clearly, is this patch available
from 2.4.20-rc1 at every default linux kernel from this moment on?
On Mon, May 19, 2003 at 03:00:47AM +0300, Halil Demirezen wrote:
> yes that is interesting, however, what i want to learn, clearly, is
> this patch available from 2.4.20-rc1 at every default linux kernel
> from this moment on?
It (ExecShield0 is *not* in the mainline kernel at this time. I don't
know when (or even if) it'll be added to the mainline kernels either.
Red Hat's Rawhide kernels (2.4.20-1.1990 for example) seem to have it,
but those are experimental kernels that are not for production use. I
guess time will tell whether Red Hat ships ExecShield by default in
their next release.
BTW, another option is the pageexec ("PaX") patch:
http://pageexec.virtualave.net/ (warning: this page has pop-up windows)
and that's integrated into a more comprehensive security patch,
grsecurity:
http://www.grsecurity.net/
-Barry K. Nathan <[email protected]>
Check out [http://www.trusteddebian.org/]. They apply a bunch of kernel
patches like grsecurity. Seems they have to recompile, repackage all
the .debs though.
It is interesting to note that some of these patches introduce an
apparent non-determinism into userspace; the same inputs could result in
different outputs (the difference is that there are more inputs than
userspace can actually see). So, maybe it could happen that you get a
segv from an off by 1, but when you rerun it, you can't reproduce it
because all the addresses are different.
In practice, this will eventually segv somewhere, and someone will find
the problem, but its something to consider.
Justin Pryzby
On Mon, May 19, 2003 at 02:22:00AM +0000, Barry K. Nathan wrote:
>
> On Mon, May 19, 2003 at 03:00:47AM +0300, Halil Demirezen wrote:
> > yes that is interesting, however, what i want to learn, clearly, is
> > this patch available from 2.4.20-rc1 at every default linux kernel
> > from this moment on?
> ...
> BTW, another option is the pageexec ('PaX') patch:
> http://pageexec.virtualave.net/ (warning: this page has pop-up windows)
>
> and that's integrated into a more comprehensive security patch,
> grsecurity:
> http://www.grsecurity.net/
>
> -Barry K. Nathan <[email protected]>