On Wed, Mar 22, 2017 at 06:56:32PM +0100, Daniel Vetter wrote: > On Wed, Mar 22, 2017 at 03:47:31PM +0200, Ville Syrjälä wrote: > > On Wed, Mar 22, 2017 at 09:36:08AM +0100, Daniel Vetter wrote: > > > To match the drm_ioctl.c we already have. > > > > > > Signed-off-by: Daniel Vetter <daniel.vetter@xxxxxxxxx> > > > --- > > > drivers/gpu/drm/drm_ioctl.c | 1 + > > > include/drm/drmP.h | 61 +------------------------- > > > include/drm/drm_ioctl.h | 102 ++++++++++++++++++++++++++++++++++++++++++++ > > > 3 files changed, 104 insertions(+), 60 deletions(-) > > > create mode 100644 include/drm/drm_ioctl.h > > > > > > diff --git a/drivers/gpu/drm/drm_ioctl.c b/drivers/gpu/drm/drm_ioctl.c > > > index 601bb0ded9d2..7f4f4f48e390 100644 > > > --- a/drivers/gpu/drm/drm_ioctl.c > > > +++ b/drivers/gpu/drm/drm_ioctl.c > > > @@ -28,6 +28,7 @@ > > > * OTHER DEALINGS IN THE SOFTWARE. > > > */ > > > > > > +#include <drm/drm_ioctl.h> > > > #include <drm/drmP.h> > > > #include <drm/drm_auth.h> > > > #include "drm_legacy.h" > > > diff --git a/include/drm/drmP.h b/include/drm/drmP.h > > > index fc2a0744413d..248d2408e56b 100644 > > > --- a/include/drm/drmP.h > > > +++ b/include/drm/drmP.h > > > @@ -80,6 +80,7 @@ > > > #include <drm/drm_file.h> > > > #include <drm/drm_debugfs.h> > > > #include <drm/drm_sysfs.h> > > > +#include <drm/drm_ioctl.h> > > > > Same question as for the debugfs thing. Why do we need this? > > Try fixing up the drmP.h include hell, and you know that it'd take a few > hundred patches. And right now it's impossible, because the most central > structure (struct drm_device) is in here, so defacto we still have a > monolithic header which needs everything. > > First we need to split it completely, then all places that include drmP.h > need to switch to individual headers (each time fixing up a pile of > fallout), and then eventually (after fixing the few hundred places or so) > we can remove all these include lines here (or well, get rid of drmP.h > entirely). But not before. I tried. OK. I thought this stuff wouldn't have spread that far beyond the ioctl implementations, but I guess not. Hmm. Had I spent any thought on this I would have realized that at least DRM_IOCTL_DEF_DRV() is used in pretty much every driver. -- Ville Syrjälä Intel OTC _______________________________________________ dri-devel mailing list dri-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/dri-devel