Re: [gentoo-hardened] ssp random bytes solution

[Robert Connolly]

On April 21, 2004 01:12 pm, pageexec@××××××××.hu wrote:
> > > 1. don't initialize __guard[] to all 0s, on some compilers it will
> > > force it into the .data section (instead of .bss), it's a slight
> > > waste...
> >
> > I have read the stack value can be used here with zero preformance loss.
> > Just to set initial values.
>
> 'stack value' = ?

I mean to say anything freely available, like the time.

> anyway, taking into account my 3rd observation as well, you should do
> something like this:

This libc patch is based on the gcc/libgcc2.c hunk of the ibm propolice patch. 
__guard is an array in there too. I also noticed its a 'static void', will 
that mess with aslr? In gentoo's ssp libc patch its not static. It looks to 
me like that is where the "#if defined(HAVE_SYSLOG)" came from. HAVE_SYSLOG 
is used durring the gcc compile, but not durring glibc compile. Someone 
copied and pasted without checking. This is in the gcc hunk but missing from 
our glibc patch (while in obsd's libc hunk):

static void __guard_setup(void) __attribute__ ((constructor));
void __stack_smash_handler(char func[], int damaged __attribute__((unused)));

Looks like I should be including sys/sysctl.h instead of linux/sysctl.h too 
(so both get included). Also, there are two ways to invoke sysctl (and 
__sysctl). The first is a wrapper, second is direct. I didn't know the 
difference so I used __sysctl.


--
gentoo-hardened@g.o mailing list