Hi Thomas, On Fri, Sep 25, 2020 at 05:01:23PM +0200, Thomas Zimmermann wrote: > >> + ARRAY_SIZE(simplekms_formats), > >> + simplekms_format_modifiers, > >> + connector); > >> + if (ret) > >> + return ret; > >> + > >> + drm_mode_config_reset(dev); > > > > This breaks fastboot. I think ideally we'd have the state represent > > everything is on, and allocate an fb + buffer with the current contents of > > the framebuffer. Since we can allocate an fb that matches this shouldn't > > be a problem, just a raw memcpy_fromio should do the job. > > I'm trying to wrap my head around how the fastboot setup is implemented. > > Apparently, i915's fbdev code goes through the existing pipeline state > and fills the fb_info structure with compatible settings. > > Where is the initial pipeline state created? If I write reset handlers > that initialize the pipeline to the simple-framebuffer's fixed state, > whould that be sufficient? A later stage could then do the equivalent of > intel_fbdev_init_bios(). > > The simple-kms helpers don't seem to support custom reset handlers or > atomic-state callbacks. I guess that would require and update as well? You probably want to read the following :) https://lore.kernel.org/dri-devel/CAKMK7uHtqHy_oz4W7F+hmp9iqp7W5Ra8CxPvJ=9BwmvfU-O0gg@xxxxxxxxxxxxxx/ It's been on my todo-list since, but I never got to work on it :/ Maxime
Attachment:
signature.asc
Description: PGP signature
_______________________________________________ dri-devel mailing list dri-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/dri-devel
