Merge branch 'linux-5.8' of git://github.com/skeggsb/linux into drm-fixes

A couple of fixes for issues relating to format modifiers (there's
still a patch pending from James Jones to hopefully address the
remaining ones), regression fix from the recent HDA nightmare, and a
race fix for Turing modesetting.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Ben Skeggs <skeggsb@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/ <CACAvsv5aAp+FZMZGTB+Nszc==h5gEbdNV58sSRRQDF1R5qQRGg@mail.gmail.com

Author: airlied

drivers/gpu/drm/nouveau/dispnv50/disp.c

@@ -2073,7 +2073,7 @@ nv50_disp_atomic_commit_tail(struct drm_atomic_state *state)
 	 */
 	if (core->assign_windows) {
 		core->func->wndw.owner(core);
-		core->func->update(core, interlock, false);
+		nv50_disp_atomic_commit_core(state, interlock);
 		core->assign_windows = false;
 		interlock[NV50_DISP_INTERLOCK_CORE] = 0;
 	}
@@ -2506,7 +2506,7 @@ nv50_display_create(struct drm_device *dev)
 	if (disp->disp->object.oclass >= TU102_DISP)
 		nouveau_display(dev)->format_modifiers = wndwc57e_modifiers;
 	else
-	if (disp->disp->object.oclass >= GF110_DISP)
+	if (drm->client.device.info.family >= NV_DEVICE_INFO_V0_FERMI)
 		nouveau_display(dev)->format_modifiers = disp90xx_modifiers;
 	else
 		nouveau_display(dev)->format_modifiers = disp50xx_modifiers;

drivers/gpu/drm/nouveau/nouveau_fbcon.c

@@ -315,7 +315,7 @@ nouveau_fbcon_create(struct drm_fb_helper *helper,
 	struct drm_framebuffer *fb;
 	struct nouveau_channel *chan;
 	struct nouveau_bo *nvbo;
-	struct drm_mode_fb_cmd2 mode_cmd;
+	struct drm_mode_fb_cmd2 mode_cmd = {};
 	int ret;
 
 	mode_cmd.width = sizes->surface_width;
@@ -590,6 +590,7 @@ nouveau_fbcon_init(struct drm_device *dev)
 	drm_fb_helper_fini(&fbcon->helper);
 free:
 	kfree(fbcon);
+	drm->fbcon = NULL;
 	return ret;
 }
 

drivers/gpu/drm/nouveau/nvkm/engine/disp/outp.c

@@ -117,15 +117,6 @@ nvkm_outp_acquire_hda(struct nvkm_outp *outp, enum nvkm_ior_type type,
 {
 	struct nvkm_ior *ior;
 
-	/* First preference is to reuse the OR that is currently armed
-	 * on HW, if any, in order to prevent unnecessary switching.
-	 */
-	list_for_each_entry(ior, &outp->disp->ior, head) {
-		if (!ior->identity && !!ior->func->hda.hpd == hda &&
-		    !ior->asy.outp && ior->arm.outp == outp)
-			return nvkm_outp_acquire_ior(outp, user, ior);
-	}
-
 	/* Failing that, a completely unused OR is the next best thing. */
 	list_for_each_entry(ior, &outp->disp->ior, head) {
 		if (!ior->identity && !!ior->func->hda.hpd == hda &&
@@ -173,6 +164,27 @@ nvkm_outp_acquire(struct nvkm_outp *outp, u8 user, bool hda)
 		return nvkm_outp_acquire_ior(outp, user, ior);
 	}
 
+	/* First preference is to reuse the OR that is currently armed
+	 * on HW, if any, in order to prevent unnecessary switching.
+	 */
+	list_for_each_entry(ior, &outp->disp->ior, head) {
+		if (!ior->identity && !ior->asy.outp && ior->arm.outp == outp) {
+			/*XXX: For various complicated reasons, we can't outright switch
+			 *     the boot-time OR on the first modeset without some fairly
+			 *     invasive changes.
+			 *
+			 *     The systems that were fixed by modifying the OR selection
+			 *     code to account for HDA support shouldn't regress here as
+			 *     the HDA-enabled ORs match the relevant output's pad macro
+			 *     index, and the firmware seems to select an OR this way.
+			 *
+			 *     This warning is to make it obvious if that proves wrong.
+			 */
+			WARN_ON(hda && !ior->func->hda.hpd);
+			return nvkm_outp_acquire_ior(outp, user, ior);
+		}
+	}
+
 	/* If we don't need HDA, first try to acquire an OR that doesn't
 	 * support it to leave free the ones that do.
 	 */