From: Jürg Billeter
Date: Tue Sep 12 2017 - 14:55:24 EST

Hi Oleg,

Thanks for the review.

On Tue, 2017-09-12 at 19:05 +0200, Oleg Nesterov wrote:
> On 09/09, JÃrg Billeter wrote:
> > Unlike
> > PR_SET_PDEATHSIG, this is inherited across fork to allow killing a whole
> > subtree without race conditions.
> but I am still not sure this is right... at least I can't understand the
> "without race conditions" above.
> IOW, the child can do prctl(PR_SET_PDEATHSIG_PROC, SIGKILL) right after fork(),
> why this is not enough to kill a whole subtree without race conditions?

What if the parent dies between fork() and prctl()? Besides avoiding
this race condition, it also makes it relatively easy to enforce
PDEATHSIG_PROC for all descendants of a process. You simply set
PDEATHSIG_PROC and then block further changes using seccomp (and set
no_new_privs) to avoid runaway children.

> OTOH. If you want to kill a whole sub-tree then perhaps the exiting process
> should simply send the ->pdeath_signal_proc to the whole sub-tree? Not that
> I really think this makes more sense, but if we add the new API we should
> discuss everything we can.

While this would likely work for my use case of avoiding runaway
processes, I don't think it would make sense for non-SIGKILL use cases
of cooperating processes. Inheritance across fork still allows
resetting PDEATHSIG_PROC in the child after fork and I don't expect the
parent death race to be a significant issue in the case of cooperating

> Say, CLONE_PARENT. Should it succeed if ->pdeath_signal_proc != 0 ?

Yes, I don't see an issue with that. The new process will be a sibling
and inheriting pdeath_signal_proc seems sensible to me for this.