Re: module auto-unload

[Paul Goyette <paul%whooppee.com@localhost>]

On Sun, 15 Dec 2013, Christos Zoulas wrote:

> > +       if (t < 0)
> > +               return (EINVAL);
>
> You are not allowing it to become negative.
>
> >  *      Automatically unload modules.  We try once to unload autoloaded
> >  *      modules after module_autotime seconds.  If the system is under
> > - *     severe memory pressure, we'll try unloading all modules.
> > + *     severe memory pressure, we'll try unloading all modules, else if
> > + *     module_autotime is non-positive, we don't try to unload.
> >  */
> > static void
> > module_thread(void *cookie)
> > @@ -1311,6 +1341,8 @@ module_thread(void *cookie)
> >
> >                         if (uvmexp.free < uvmexp.freemin) {
> >                                 module_thread_ticks = hz;
> > +                       } else if (module_autotime <= 0) {
> > +                               continue;
>
> Yet, you add a check for it. I would just not add the two lines above,
> and have only 0 mean "never unload"

Yes, this check could be simply == instead of <=

I left this check here, though, so it would not auto-unload any modules 
that were loaded before setting module_autotime to 0.  In other words,

        T=0    module_autotime == 10 (the default)
        T=1    X is autoloaded, and is scheduled for auto-unload
        T=2    module_autotime is set to 0
        T=3    module_thread() kicks in and unloads X

With the check-for-0 in module_thread() we avoid this unload.  I could 
live with it either way, but it seemed more logical to me to avoid all 
auto-unloads as soon as the time-out was set to zero, rather than only 
avoiding timeouts for modules that were loaded after the set-to-zero.


I have just completed testing this change, and it seems to work just 
fine, at least on my amd64 systems.  Is there any reason why these 
changes should not be committed?




-------------------------------------------------------------------------
| Paul Goyette     | PGP Key fingerprint:     | E-mail addresses:       |
| Customer Service | FA29 0E3B 35AF E8AE 6651 | paul at whooppee.com    |
| Network Engineer | 0786 F758 55DE 53BA 7731 | pgoyette at juniper.net |
| Kernel Developer |                          | pgoyette at netbsd.org  |
-------------------------------------------------------------------------