Re: [PATCH] audit: add arch field to seccomp event log

[Richard Guy Briggs]

On 14/02/18, Steve Grubb wrote:
> On Tuesday, February 18, 2014 03:50:44 PM Richard Guy Briggs wrote:
> > > missing '='   but this isn't what audit_get_context() does...   it's
> > > crappy naming...    I'd think a combo of audit_dummy_context() and
> > > current->audit_context would be most appropriate.
> > 
> > Ok.  I think I finally understand audit_dummy_context().  Thanks for the
> > hint.  However, it appears it is not useful in this sitation, since if
> > there is an audit_context, even a dummy context, it appears arch is
> > filled in.
> > 
> > diff --git a/kernel/auditsc.c b/kernel/auditsc.c
> > @@ -2406,12 +2406,18 @@ void audit_core_dumps(long signr)
> >  void __audit_seccomp(unsigned long syscall, long signr, int code)
> >  {
> >         struct audit_buffer *ab;
> > +       struct audit_context *context = current->audit_context;
> >  
> >         ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_SECCOMP);
> >         if (unlikely(!ab))
> >                 return;
> >         audit_log_task(ab);
> >         audit_log_format(ab, " sig=%ld", signr);
> > +       audit_log_format(ab, " arch=");
> > +       if (context)
> > +               audit_log_format(ab, "%x", context->arch);
> > +       else
> > +               audit_log_format(ab, "(null)");
> >         audit_log_format(ab, " syscall=%ld", syscall);
> >         audit_log_format(ab, " compat=%d", is_compat_task());
> >         audit_log_format(ab, " ip=0x%lx", KSTK_EIP(current));
> 
> Is there anything that could be passed by the caller that might identify the 
> syscall ABI when this call was blocked? '(null)' still makes syscall number 
> uninterpretable.

This is the way it is done in the existing record that prints a syscall
record in audit_log_exit().

It is set by the arch-dependent assembler linkage for
__audit_syscall_entry() or set by the arch-dependent ptrace code that
calls audit_syscall_entry().  In some arches there are a couple of
choices.


To pass this in to __audit_seccomp() as an arg, I'd have to add that arg
to:
	__audit_seccomp()
	audit_seccomp()
	__secure_computing()
	secure_computing()


As you've pointed out in IRC, there is a syscall_get_arch(current, regs)
that could be used to get the arch when an audit context is not yet created.

I was just looking for something like that...  I am noticing that all
arches call a variant of syscall_trace_enter() which first calls
secure_computing() [which eventually calls __audit_seccomp()] *before*
calling audit_syscall_entry() [which eventually sets context->arch].  So
I'm not sure how task->audit_context->arch got set...  Perhaps
__audit_syscall_entry() should just call syscall_get_arch() instead of
lugging it through the stack like it presently does.

Is this going to give the same information?  (I guess I should be able
to answer that question...)

> -Steve

- RGB

--
Richard Guy Briggs <rbriggs@xxxxxxxxxx>
Senior Software Engineer, Kernel Security, AMER ENG Base Operating Systems, Red Hat
Remote, Ottawa, Canada
Voice: +1.647.777.2635, Internal: (81) 32635, Alt: +1.613.693.0684x3545
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/