It's probably full of bugs ready for exploiting by userspace. And
there's not going to be any userspace for this without any of the drm
legacy drivers enabled too. So just couple it together.
Note that the frontend is only the /dev/agp ioctl interface, which per
Adam is only used by the i810 userspace drivers. All other drivers go
through the drm bufmap agp handling abstraction apparently.
v2: Augment commit message a bit from m-l feedback.
Acked-by: Adam Jackson <ajax@redhat.com>
Acked-by: Christian König <christian.koenig@amd.com>
Acked-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Cc: David Airlie <airlied@linux.ie>
Cc: Adam Jackson <ajax@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201117214029.591896-1-daniel.vetter@ffwll.ch
# SPDX-License-Identifier: GPL-2.0
-agpgart-y := backend.o frontend.o generic.o isoch.o
+agpgart-y := backend.o generic.o isoch.o
+ifeq ($(CONFIG_DRM_LEGACY),y)
agpgart-$(CONFIG_COMPAT) += compat_ioctl.o
+agpgart-y += frontend.o
+endif
+
obj-$(CONFIG_AGP) += agpgart.o
obj-$(CONFIG_AGP_ALI) += ali-agp.o
void agp_remove_bridge(struct agp_bridge_data *bridge);
/* Frontend routines. */
+#if IS_ENABLED(CONFIG_DRM_LEGACY)
int agp_frontend_initialize(void);
void agp_frontend_cleanup(void);
+#else
+static inline int agp_frontend_initialize(void) { return 0; }
+static inline void agp_frontend_cleanup(void) {}
+#endif
/* Generic routines. */
void agp_generic_enable(struct agp_bridge_data *bridge, u32 mode);