svn commit: r336326 - head/sys/dev/ichiic

[Warner Losh]

On Mon, Jul 16, 2018, 10:57 AM John Baldwin <jhb at freebsd.org> wrote:

> On 7/15/18 6:34 PM, Oleksandr Tymoshenko wrote:
> > Author: gonzo
> > Date: Mon Jul 16 01:34:45 2018
> > New Revision: 336326
> > URL: https://svnweb.freebsd.org/changeset/base/336326
> >
> > Log:
> >   Remove MODULE_PNP_INFO for ig4(4) driver
> >
> >   ig4(4) does not support suspend/resume but present on the hardware
> where
> >   such functionality is critical, like laptops. Remove PNP info to avoid
> >   breaking suspend/resume on the systems where ig4(4) load is not
> explicitly
> >   requested by the user.
> >
> >   PR:             229791
> >   Reported by:    Ali Abdallah
>
> I wonder if we don't want instead a way to tag a module as "don't autoload"
> so that you can still have a way to say "show me which unprobed devices
> have modules available".  One option would be to have an rc.conf knob which
> is a regex that gets used with 'egrep -v' in the rc.d script on the output
> of devmatch.  This has the slight advantage that a user can customize it
> via rc.conf.  (For example, when working on cxgbe(4) on my dev boxes I
> don't
> want it to autoload, but in a production environment we might want it to
> autoload by default.)  Another option might be to add some new type of
> PNP info that is a boolean "autoload" that defaults to true and teach
> devmatch to honor the boolean.
>

How does this break suspend / resume? That makes no sense to me at all. I
think if this changes anything, it only we could fix that by accident.

I get wanting to have a blacklist, but that belongs in rc, not the driver.
That's on my soon list. But I need to implement the freeze thaw of proving
first so I can re.ove the ugly kludge for usb first so we can filter again.

We also need a newbus flag that says the device has been attached at least
once so kldunload works...

Warner

>