Re: [RFC PATCH 1/3] seccomp: Don't allow tracers to abuse RET_TRACE

["Indan Zupancic" <indan@nul.nu>]

On Thu, May 24, 2012 18:07, Will Drewry wrote:
> Ensure that consumers of the PTRACE_EVENT_SECCOMP notification
> cannot change the system call number for the traced task
> without it resulting in the system call being skipped.
>
> Traditionally, tracers will set the system call number to
> -1 to skip the system call. This behavior will work as expected
> but the tracer will be unable to remap the system call to a valid
> system call after the seccomp policy has been evaluated.
>
> Signed-off-by: Will Drewry <wad@chromium.org>
> ---
>  kernel/seccomp.c |    4 ++++
>  1 file changed, 4 insertions(+)
>
> diff --git a/kernel/seccomp.c b/kernel/seccomp.c
> index ee376be..33f0ad6 100644
> --- a/kernel/seccomp.c
> +++ b/kernel/seccomp.c
> @@ -425,6 +425,10 @@ int __secure_computing(int this_syscall)
>  			 */
>  			if (fatal_signal_pending(current))
>  				break;
> +			/* Skip the system call if the tracer changed it. */
> +			if (this_syscall !=
> +			    syscall_get_nr(current, task_pt_regs(current)))
> +				goto skip;
>  			return 0;
>  		case SECCOMP_RET_ALLOW:
>  			return 0;
> --

This patch doesn't make any sense whatsoever. You can't know why a system
call was blocked by a seccomp filter, assuming it's always because of the
system call number is wrong.

Also, you don't check if an allowed system call is changed into a denied
one, so this doesn't protect against ptracers bypassing seccomp filters.

And one of the main points of PTRACE_EVENT_SECCOMP events was that it's
useful for cases that can't be handled or decided by the seccomp filter.
Then taking away the ability to change the syscall number makes it a lot
less useful.

Either do the seccomp test before or after ptrace, or both, but please
don't introduce ad hoc checks like this.

Greetings,

Indan