Re: PT_EXITKILL (Was: pdeath_signal)

[Pedro Alves <palves@redhat.com>]

On 11/07/2012 03:09 PM, Oleg Nesterov wrote:

>> > > > > What I would IDEALLY like to have is a call, probably a ptrace option,
>> > > > > where the parent can request: "If I am ever to terminate or be killed,
>> > > > > then my ptraced son MUST die as well".
>> > >
>> > > Perhaps this makes sense...
>> > >
>> > > Chris, iirc you also suggested something like this? And the patch is
>> > > trivial.

(...)

> OK. Please see the untested/uncompiled (but trivial) patch below
> 
> 	- it adds PTRACE_O_EXITKILL. A better name?
> 
> 	- A better numeric value? Note that the new option is not equal to
> 	  the last-ptrace-option << 1. Because currently all options have
> 	  the event, and the new one starts the eventless group. 1 << 16
> 	  means we have the room for 8 more events.
> 
> 	- it needs the convincing changelog for akpm


If this isn't inherited by the ptrace child's children, a fork child can
end up detached if the tracer dies before it had a chance of setting
the PTRACE_O_EXITKILL on the new auto-attached child.

Which sounds like another argument for PTRACE_O_INHERIT, as in:
http://sourceware.org/ml/archer/2011-q1/msg00026.html

(it sounds like you need to use PTRACE_SEIZE+options too to plug
the race between PTRACE_ME/PTRACE_ATTACH and
setting PTRACE_SETOPTIONS).

(For completeness, Windows' age old equivalent,
DebugSetProcessKillOnExit, it a tracer option, not tracee option, though
that's not as flexible.)

-- 
Pedro Alves