On Sat, Dec 14, 2013 at 12:11:01AM +0200, Aleksandr Rybalko wrote: > Konstantin Belousov <kostik...@gmail.com> написав(ла): > >On Thu, Dec 12, 2013 at 02:49:27PM +0000, Aleksandr Rybalko wrote: > >> Author: ray > >> Date: Thu Dec 12 14:49:26 2013 > >> New Revision: 259261 > >> URL: http://svnweb.freebsd.org/changeset/base/259261 > >> > >> Log: > >> Do not try to probe/attach if attempt to add fbd child are failed. > >> > >> Sponsored by: The FreeBSD Foundation > >> > >> Modified: > >> head/sys/dev/drm2/drm_fb_helper.c > >> > >> Modified: head/sys/dev/drm2/drm_fb_helper.c > >> > >============================================================================== > >> --- head/sys/dev/drm2/drm_fb_helper.c Thu Dec 12 14:37:25 > >2013 (r259260) > >> +++ head/sys/dev/drm2/drm_fb_helper.c Thu Dec 12 14:49:26 > >2013 (r259261) > >> @@ -1043,7 +1043,10 @@ int drm_fb_helper_single_fb_probe(struct > >> > >> kdev = fb_helper->dev->device; > >> fbd = device_add_child(kdev, "fbd", device_get_unit(kdev)); > >> - ret = device_probe_and_attach(fbd); > >> + if (fbd != NULL) > >> + ret = device_probe_and_attach(fbd); > >> + else > >> + ret = ENODEV; > >This must be -ENODEV, since linux layer operates on negative values for > >error. The error is negated in the drm layer, which would result in > >non-existent errno returned to FreeBSD code. > > It is not actually return code, but only value to show in error message few > lines later. ENODEV just value for case no device attached. > That error message appear only if kernel configuration include device vt. > (planning for syscons too, maybe).
Ok.
I looked around some more, and I think that the call to
drm_fb_helper_set_par() is required to properly initialize crtc.
See the #ifdef-ed Linux code.
>
> >
> >Also, the #ifdef __FreeBSD braces in the file are not useful, please
> >remove them.
> >> #ifdef DEV_VT
> >> if (ret != 0)
> >> DRM_ERROR("Failed to attach fbd device: %d\n", ret);
>
> Should we with Jean have to remove all #ifdef __FreeBSD and __Linux over
> whole DRM2 code, to reduce differences with linux code?
>
The #ifdefs should be removed.
The removal would not reduce difference with the upstream, but this is not
the point anyway.
pgpo3G__CQnME.pgp
Description: PGP signature
