@@ -1369,27 +1362,57 @@ int drm_fb_helper_setcmap(struct fb_cmap *cmap, struct fb_info *info) memcpy(g + cmap->start, cmap->green, cmap->len * sizeof(*g)); memcpy(b + cmap->start, cmap->blue, cmap->len * sizeof(*b));
for (j = 0; j < cmap->len; j++) {u16 hred, hgreen, hblue, htransp = 0xffff;
ret = crtc->funcs->gamma_set(crtc, r, g, b,crtc->gamma_size, crtc_state);I guess my description of what I have in mind wasn't really clear. I think a proper atomic commit should never reuse one of the old hooks (->gamma_set) here, that's just confusing. Instead what I had in mind is to do the proper adjusting that gamma_set does here in this function, i.e.
create the new blob, fill it with the cmap data
assign that blob to the crtc state:
drm_atomic_replace_property_blob(&crtc_state->gamma_lut, new_table, &temp);
That function is static, and...
Missed these comments here. I think we should just make them unstatic, that might also explain why gamma_set is going this indirect path. Means we need some kerneldoc, but that's not much work. Ping the original author if you have questions.
Cheers, Daniel