cvs commit: src/sys/kern kern_sysctl.c

[Maxime Henrion]

John Baldwin wrote:
> 
> On 29-May-2003 Maxime Henrion wrote:
> > Nate Lawson wrote:
> >> On Thu, 29 May 2003, Maxime Henrion wrote:
> >> >   Modified files:
> >> >     sys/kern             kern_sysctl.c
> >> >   Log:
> >> >   When loading a module that contains a sysctl which is already compiled
> >> >   in the kernel, the sysctl_register() call would fail, as expected.
> >> >   However, when unloading this module again, the kernel would then panic
> >> >   in sysctl_unregister().  Print a message error instead.
> >> >
> >> > +  /*
> >> > +   * This can happen when a module fails to register and is
> >> > +   * being unloaded afterwards.  It should not be a panic()
> >> > +   * for normal use.
> >> > +   */
> >> > +  if (error)
> >> > +          printf("%s: failed to unregister sysctl\n", __func__);
> >> 
> >> Thank you, this is helpful.  However, we have quite a few error messages
> >> that appear when an attach fails.  Is this one necessary in practice or
> >> should the eventual plan be to change the API to return an errno?  (Note,
> >> not talking about 5.1 here).
> > 
> > I agree with you that we should think about this issue a bit more once
> > 5.1 is out, as I'm not comfortable with this error message.  However, if
> > I remember right, changing the API is not an option after 5.1 since we
> > promised to not break APIs and ABIs past this release.  If the message
> > turns out to be annoying, it can be removed a bit later and we can have
> > sysctl_unregister() silently fail.  Changing the API would probably be
> > better, but I bet we'll have to do this for 6.0.
> 
> AFAIK, the API and ABI is not frozen until RELENG_5 is branched.
> That isn't happening at 5.1 release, so there is still some time to
> fix the API/ABI if need be.  We really should start avoiding making
> API/ABI changes though after 5.1 is out.

Ok, thanks for the clarification!

Maxime