Disabling ptrace

[Konstantin Belousov]

On Tue, Dec 30, 2014 at 03:07:10PM +0100, Jilles Tjoelker wrote:
> On Tue, Dec 30, 2014 at 01:19:41PM +0200, Konstantin Belousov wrote:
> > The question about a facility to disable introspection functionality
> > (ptrace etc) for a process was asked several times. The latest query
> > made me actually code the feature. Note that other systems, e.g. Linux
> > and OSX, do have similar facilities.
> 
> > Patch is below, it provides two new procctl(2) requests.
> > PROC_TRACE_ENABLE enables or disables tracing.  It includes core
> > dumping, ptrace, ktrace, debugging sysctls and hwpmc.
> > PROC_TRACE_STATUS allows to get the tracing state.
> 
> > Most interesting question is how should disabling of trace behave
> > with regard of fork and exec. IMO, the right model is to protect
> > access to the _program_ address space, which translates to inheritance
> > of the attribute for fork, and reenabling the tracing on exec.
> 
> I agree. I imagine this will be useful for programs like ssh-agent, to
> protect their unlocked key material.
> 
> This is also what Linux provides, and it is simpler than this patch:
> prctl(PR_SET_DUMPABLE) lets a process make their issetugid() equivalent
> return true, including preventing tracing by unprivileged users. You
> could call that unification a hack.
Yes, I do not like this.  We have nice and proper p_candebug(9) KPI.

> 
> > On the other hand, I understand that some users want to inherit the
> > tracing disable on exec, so there are PROC_TRACE_SET_DISABLED and
> > PROC_TRACE_SET_DISABLED_EXEC, the later makes disable to be kept after
> > exec.
> 
> This is apparently meant to protect a whole process tree as a hardening
> measure, or instead of PROC_TRACE_SET_DISABLED if it is undesirable to
> modify the program with key material.
Agreed, it could be reinterpreted this way.  Do you suggest to change
name for PROC_TRACE_SET_DISABLED_EXEC ?  E.g. PROC_TRACE_SET_DISABLED_TREE ?

> 
> > Note that it is trivial for root on the host to circumvent the feature.
> 
> I'd prefer if root can still trace normally, without needing any hacks.
> Philosophically, FreeBSD should serve the system administrator first and
> only then the application programmer. Also, the debugging facilities may
> be needed to debug FreeBSD itself (e.g. procstat -k), not just the
> application.

This is reasonable. It seems that the only way to enable host root to
use tracing without allowing jail' roots to do the same, is to introduce
new privilege. I changed the p_candebug() chunk to the following:

	/* Denied explicitely */
	if ((p->p_flag2 & P2_NOTRACE) != 0) {
		error = priv_check(td, PRIV_DEBUG_DENIED);
		if (error != 0)
			return (error);
	}