Re: [RFC][PATCH] HWPOISON: only early kill processes who installedSIGBUS handler

[Wu Fengguang]

On Wed, Jun 17, 2009 at 06:00:06PM +0800, Nick Piggin wrote:
> On Wed, Jun 17, 2009 at 05:55:32PM +0800, Wu Fengguang wrote:
> > On Wed, Jun 17, 2009 at 04:04:04PM +0800, Nick Piggin wrote:
> > > Well then you can still early-kill random apps that did not
> > > want it, and you may still cause problems if its sigbus
> > > handler does something nontrivial.
> > > 
> > > Can you use a prctl or something so it can expclitly
> > > register interest in this?
> > 
> > No I don't think prctl would be much better.
> > 
> > - if an application want early/late kill, it can do so with a proper
> >   written SIGBUS handler: the prctl call is redundant.
> 
> s/proper written/is switched to new semantics based on the existance
> of a/

Not necessarily so. If an application
- did not has a SIGBUS handler, and want to be
  - early killed: must install a handler, this is not a big problem
    because it may well want to rescue something on the event.
  - late killed: just do nothing.
(here kill = 'notification')
- had a SIGBUS hander, and want to
  - early die: call exit(0) in the handler.
  - late die: intercept and ignore the signal.
So if source code modification is viable, prctl is not necessary at all.

> > - if an admin want to control early/late kill for an unmodified app,
> >   prctl is as unhelpful as this patch(*).
> 
> Clearly you can execute a process with a given prctl.

OK, right.

> > - prctl does can help legacy apps whose SIGBUS handler has trouble
> >   with the new SIGBUS codes, however such application should be rare
> >   and the application should be fixed(why shall it do something wrong
> >   on newly introduced code at all? Shall we stop introducing new codes
> >   just because some random buggy app cannot handle new codes?)
> 
> Backwards compatibility? Kind of important.

Maybe.

> > So I still prefer this patch, until we come up with some solution that
> > allows both app and admin to change the setting.
> 
> Not only does it allow that, but it also provides backwards
> compatibility. Your patch does not allow admin to change
> anything nor does it guarantee 100% back compat so I can't
> see how you think it is better.

I didn't say it is better, but clearly mean that prctl is not better
enough to warrant a new user interface, if(!adm_friendly). Now it's
obvious that adm_friendly=1, so I agree prctl is a good interface :)

> Also it does not allow for an app with a SIGBUS handler to
> use late kill. If late kill is useful to anyone, why would
> it not be useful to some app with a SIGBUS handler (that is
> not KVM)?

Late kill will always be sent. Ignore the early kill signal in the
SIGBUS handler does the trick (see above analyzes).

Thanks,
Fengguang
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/