svn commit: r296320 - in head/sys: kern sys

[Konstantin Belousov]

On Thu, Mar 03, 2016 at 01:12:05PM +0100, Hans Petter Selasky wrote:
> On 03/03/16 12:58, Konstantin Belousov wrote:
> >> >
> >> >Yes, your change is compatible if you compile the code, but not binary
> >> >compatible. See the following macros, and how they use _callout_stop_safe():
> > My change is binary compatible.
> >
> >> >
> >> >#define callout_stop(c)         _callout_stop_safe(c, 0, NULL)
> >> >#define callout_async_drain(c, d)                                       \
> >> >      _callout_stop_safe(c, 0, d)
> >> >
> >> >Because callout_stop(c) is a macro, you've now change the API for
> >> >callout_stop() aswell. This is what I think will break binary
> >> >compatibility in 10.x.
> > How my commit changes anything for either callout_stop(), or for
> > callout_async_drain(), or for callout_drain() ?
> >
> 
> Hi,
> 
> It looked like you added another argument to _callout_stop_safe(). After 
> a closer look I see the patch is binary compatible.

Right, and CS_DRAIN value was selected so that callout_drain() arguments
to _callout_stop_safe() would be same with or without the change.

The only case where the issue could exists, is for some code which called
_callout_stop_safe() directly, and passed arbitrary (not "1") non-zero
value as the second argument to indicate drain.  I believe there is
no such code, and tend to consider it as bug, if demostrated.
be not different with