where is device_get_parent function defined

[Barney Cordoba]

--- On Wed, 9/23/09, Andriy Gapon <avg at icyb.net.ua> wrote:

> From: Andriy Gapon <avg at icyb.net.ua>
> Subject: Re: where is device_get_parent function defined
> To: "Stefan Ehmann" <shoesoft at gmx.net>
> Cc: freebsd-current at freebsd.org, "Barney Cordoba" <barney_cordoba at yahoo.com>, "pluknet" <pluknet at gmail.com>
> Date: Wednesday, September 23, 2009, 8:00 AM
> on 22/09/2009 23:44 Stefan Ehmann
> said the following:
> > On Tuesday 22 September 2009 16:21:54 Barney Cordoba
> wrote:
> >> --- On Tue, 9/22/09, pluknet <pluknet at gmail.com>
> wrote:
> > ...
> >>> Following style(9):
> >>> ###
> >>>      The function type should
> be on a
> >>> line by itself preceding the function.
> >>>
> >>>      static char *
> >>>      function(int a1, int a2,
> float fl,
> >>> int a4)
> >>> ###
> >>>
> >>> So you can safely use the caret sign in regex:
> grep
> >>> ^keyword path
> >> Except for the 50K recursive warnings from the
> module build directories
> >> you're correct.
> > 
> > In that case, I normally use:
> > find /usr/src/ -type f | xargs grep
> ^device_get_parent
> > 
> > Additionally specifying -name '*.c' should even be
> faster.
> > 
> 
> I am surprised with this whole 'recursive' issue because in
> a clean source tree
> there are no recursive symlinks. They only come to be if a
> developer forgets to
> run 'make obj' before doing 'make' in modules directories.
> I think that it is a good practice to never pollute the
> source tree with build
> objects, they should belong to /usr/obj or equivalent.

If you make modules from the module directory (rather than a full
make in the kernel build directory) then you are not using the exact 
same switches, as I understand it. I had the problem with a witness 
kernel and it was quite a disaster.

barney