Re: [PATCH 13/20] signal: Implement force_fatal_sig

[Andy Lutomirski]

On 10/20/21 13:05, Linus Torvalds wrote:
> On Wed, Oct 20, 2021 at 7:45 AM Eric W. Biederman <ebiederm@xxxxxxxxxxxx> wrote:
> > Add a simple helper force_fatal_sig that causes a signal to be
> > delivered to a process as if the signal handler was set to SIG_DFL.
> >
> > Reimplement force_sigsegv based upon this new helper.
>
> Can you just make the old force_sigsegv() go away? The odd special
> casing of SIGSEGV was odd to begin with, I think everybody really just
> wanted this new "force_fatal_sig()" and allow any signal - not making
> SIGSEGV special.

I'm rather nervous about all this, and I'm also nervous about the 
existing code.  A quick skim is finding plenty of code paths that assume 
force_sigsegv (or a do_exit that this series touches) are genuinely 
unrecoverable.  For example:

- rseq: the *kernel* will be fine if a signal is handled, but the 
userspace process may be in a very strange state.

- bprm_execve: The comment says it best:

        /*
         * If past the point of no return ensure the code never
         * returns to the userspace process.  Use an existing fatal
         * signal if present otherwise terminate the process with
         * SIGSEGV.
         */
        if (bprm->point_of_no_return && !fatal_signal_pending(current))
                force_sigsegv(SIGSEGV);

- vm86: already discussed

Now force_sigsegv() at least tries to kill the task, but not very well. 
With the whole series applied and force_sigsegv() gone, these errors 
become handleable, and that needs real care.

(I don't think bprm_execve() is exploitable.  It looks like it's 
attackable in the window between setting point_of_no_return and 
unshare_sighand(), but I'm not seeing any useful way to attack it unless 
a core dump is already in progress or a *different* fatal signal is 
already pending, and in either of those cases we're fine.)