Re: [PATCH] ptrace: make former thread ID available via PTRACE_GETEVENTMSG after PTRACE_EVENT_EXEC stop (v.2)

[Tejun Heo <tj@kernel.org>]

Hello,

On Tue, Jun 28, 2011 at 02:30:36PM +0200, Denys Vlasenko wrote:
> On Tue, Jun 28, 2011 at 10:25 AM, Tejun Heo <tj@kernel.org> wrote:
> > On Mon, Jun 27, 2011 at 05:18:27PM +0200, Oleg Nesterov wrote:
> > Yeah, but that's a pretty silly way to do it.  If we make it depend on
> > PT_SEIZED, we can simply say "if seized, EXEC reports..." but as it
> > currently stands, it would go like "If the message is non-zero on
> > EXEC, it indicates...  This behavior is valid since kernel version
> > x.x.x".
> 
> This is true for any new addition to API.
> It starts from some kernel version.

Hmmm... but as I wrote above, we have a choice to make here and the
two options are clearly different?

> > Maybe adding a guarantee that PTRACE_SEIZE capable kernel
> > always reports the old pid on EXEC but that would still seem
> > unnecessarily complicated.  It isn't a bug fix.  I don't see much
> > point in introducing new behavior separately.
> 
> This new feature looks orthogonal to PTRACE_SEIZE to me.

Yeah, sure, but we're bundling a number of behavior changes with
PT_SEIZED because it's impractical to introduce each with its own
compatibility / feature test bits, so it makes sense to do the same
with this one, I think.  It's not like there's a lot to gain by
supporting this outside of SEIZE anyway.

Thanks.

-- 
tejun