<div dir="ltr">Would kms_swrast + vgem for surfaceless also work for the use case raised in the bug, or is that a no-go because it assumes the presence of a driver? I've been testing it[1], it works pretty well, except for one issue during ChromeOS startup.<div class="gmail_extra"> </div><div class="gmail_extra">[1] <a href="https://chromium-review.googlesource.com/c/558218">https://chromium-review.googlesource.com/c/558218</a></div><div class="gmail_extra"> <div class="gmail_quote">On Wed, Aug 2, 2017 at 4:05 AM, Emil Velikov <<a href="mailto:emil.l.velikov@gmail.com" target="_blank">emil.l.velikov@gmail.com</a>> wrote: <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">[+Chad] Hi Akihiko Odaki, Thank you for the patch and welcome to Mesa! How you tested this patch? Did you run a test suite like piglit, dEQP, etc? Can you give it a spin with either one of these, comparing HW vs swrast surfacess and share the results. On 1 August 2017 at 06:49, Akihiko Odaki <<a href="mailto:nekomanma@pixiv.co.jp" target="_blank">nekomanma@pixiv.co.jp</a>> wrote: > Bugzilla: <a href="https://bugs.freedesktop.org/show_bug.cgi?id=101397" rel="noreferrer" target="_blank">https://bugs.freedesktop.org/show_bug.cgi?id=101397</a> > --- > src/egl/drivers/dri2/platform_surfaceless.c | 114 ++++++++++++++++++++++++++-- > src/gallium/state_trackers/dri/drisw.c | 45 ++++++++++- A general rule of thumb - the difference in paths indicate these two are distinct components. As such they ought to be separate patches. But before that, please look at my other comment below. <div><div class="gmail-m_-184050830769618726h5"> > 2 files changed, 148 insertions(+), 11 deletions(-) > > diff --git a/src/egl/drivers/dri2/platform_surfaceless.c b/src/egl/drivers/dri2/platform_surfaceless.c > index 1091b4febd..5487c89816 100644 > --- a/src/egl/drivers/dri2/platform_surfaceless.c > +++ b/src/egl/drivers/dri2/platform_surfaceless.c > @@ -133,9 +133,16 @@ dri2_surfaceless_create_surface(_EGLDriver *drv, _EGLDisplay *disp, EGLint type, > if (!config) > goto cleanup_surface; > > - dri2_surf->dri_drawable = > - dri2_dpy->image_driver->createNewDrawable(dri2_dpy->dri_screen, config, > - dri2_surf); > + if (dri2_dpy->image_driver) { > + dri2_surf->dri_drawable = > + dri2_dpy->image_driver->createNewDrawable(dri2_dpy->dri_screen, config, > + dri2_surf); > + } else { > + assert(dri2_dpy->swrast); > + dri2_surf->dri_drawable = > + dri2_dpy->swrast->createNewDrawable(dri2_dpy->dri_screen, config, > + dri2_surf); > + } > if (dri2_surf->dri_drawable == NULL) { > _eglError(EGL_BAD_ALLOC, "image->createNewDrawable"); > goto cleanup_surface; > @@ -229,7 +236,24 @@ surfaceless_add_configs_for_visuals(_EGLDriver *drv, _EGLDisplay *dpy) > return (config_count != 0); > } > > -static const struct dri2_egl_display_vtbl dri2_surfaceless_display_vtbl = { > +static const struct dri2_egl_display_vtbl dri2_surfaceless_swrast_display_vtbl = { > + .create_pixmap_surface = dri2_fallback_create_pixmap_surface, > + .create_pbuffer_surface = dri2_surfaceless_create_pbuffer_surface, > + .destroy_surface = surfaceless_destroy_surface, > + .create_image = dri2_create_image_khr, > + .swap_interval = dri2_fallback_swap_interval, > + .swap_buffers = surfaceless_swap_buffers, > + .swap_buffers_region = dri2_fallback_swap_buffers_region, > + .set_damage_region = dri2_fallback_set_damage_region, > + .post_sub_buffer = dri2_fallback_post_sub_buffer, > + .copy_buffers = dri2_fallback_copy_buffers, > + .query_buffer_age = dri2_fallback_query_buffer_age, > + .create_wayland_buffer_from_image = dri2_fallback_create_wayland_buffer_from_image, > + .get_sync_values = dri2_fallback_get_sync_values, > + .get_dri_drawable = dri2_surface_get_dri_drawable, > +}; > + > +static const struct dri2_egl_display_vtbl dri2_surfaceless_dri3_display_vtbl = { > .create_pixmap_surface = dri2_fallback_create_pixmap_surface, > .create_pbuffer_surface = dri2_surfaceless_create_pbuffer_surface, > .destroy_surface = surfaceless_destroy_surface, > @@ -252,6 +276,66 @@ surfaceless_flush_front_buffer(__DRIdrawable *driDrawable, void *loaderPrivate) > { > } > > +static const __DRIextension *swrast_loader_extensions[] = { > + NULL, </div></div>I'm not 100% sure if it makes sense, for swrast at least, to have no loader extensions. Need to think about this a bit more. Although a related series comes to mind [1]. It adds partial DRI_IMAGE support to state_trackers/dri/drisw. I'm wondering if we cannot reuse/share some goals across the board. Thanks Emil [1] <a href="https://patchwork.freedesktop.org/project/mesa/patches/?submitter=16100&state=&q=&archive=&delegate=" rel="noreferrer" target="_blank">https://patchwork.freedesktop.org/project/mesa/patches/?submitter=16100&state=&q=&archive=&delegate=</a> <div class="gmail-m_-184050830769618726HOEnZb"><div class="gmail-m_-184050830769618726h5">_______________________________________________ mesa-dev mailing list <a href="mailto:mesa-dev@lists.freedesktop.org" target="_blank">mesa-dev@lists.freedesktop.org</a> <a href="https://lists.freedesktop.org/mailman/listinfo/mesa-dev" rel="noreferrer" target="_blank">https://lists.freedesktop.org/mailman/listinfo/mesa-dev</a> </div></div></blockquote></div> </div></div>