On Wed, Dec 13, 2023 at 02:59:07AM +0200, Ville Syrjälä wrote: > On Wed, Dec 13, 2023 at 02:42:26AM +0200, Ville Syrjala wrote: > > diff --git a/drivers/gpu/drm/i915/display/intel_fbdev_fb.c b/drivers/gpu/drm/i915/display/intel_fbdev_fb.c > > index 717c3a3237c4..1ac05d90b2e8 100644 > > --- a/drivers/gpu/drm/i915/display/intel_fbdev_fb.c > > +++ b/drivers/gpu/drm/i915/display/intel_fbdev_fb.c > > @@ -78,7 +78,7 @@ int intel_fbdev_fb_fill_info(struct drm_i915_private *i915, struct fb_info *info > > > > /* Use fbdev's framebuffer from lmem for discrete */ > > info->fix.smem_start = > > - (unsigned long)(mem->io_start + > > + (unsigned long)(mem->io.start + > > i915_gem_object_get_dma_address(obj, 0)); > > Hmm. That looks wrong for MTL actually since dma address is relative > to the start of LMEMBAR but stolen io.start will be LMEMBAR+8MiB (or > just DSMBASE which points to the same place, with the w/a in place). > So we need to subtract mem->region.start from this to get the correct > value. I suppose this doesn't matter anymore as we have our own .fb_mmap() these days. So presumably we could just always leave this set to zero. -- Ville Syrjälä Intel
