panfrost patches

README.rst

@@ -1,59 +1,136 @@
 `Mesa <https://mesa3d.org>`_ - The 3D Graphics Library
 ======================================================
 
+Valhall v10 "CSF" support branch—for Mali G710/G610.
+
+Note that firmware is required for these GPUs, for RK3588 try
+downloading the file from the Rockchip `libmali
+<https://github.com/JeffyCN/rockchip_mirrors/tree/libmali/firmware/g610>`_
+repo, and placing it in ``/lib/firmware/``.
+
+Windowing system support
+------------------------
+
+Panfrost Wayland compositor (wlroots):
+
+#. Panfrost Wayland clients
+#. Panfrost X11 clients via Xwayland [1]_
+#. Blob X11 clients via Xwayland + dri2to3 [2]_
+
+Panfrost Wayland compositor (non-wlroots):
+
+#. Panfrost Wayland clients
+#. Panfrost X11 clients via Xwayland
+#. Blob Wayland clients
+#. Blob X11 clients via Xwayland + dri2to3 [2]_
+
+Blob Wayland compositor:
+
+#. Panfrost Wayland clients
+#. Blob Wayland clients
+
+Panfrost Xorg server: [3]_
+
+#. Panfrost X11 clients
+#. Blob X11 clients
+
+Blob Xorg server:
+
+#. Panfrost X11 clients
+#. Blob X11 clients
+
+Applications using KMS/DRM will also work.
+
+.. [1] Requires ``CONFIG_DRM_IGNORE_IOTCL_PERMIT`` to be disabled in
+       the kernel configuration. The option is broken and should never
+       be enabled anyway.
+
+.. [2] See https://gitlab.com/panfork/dri2to3
+
+.. [3] For Radxa Debian/Ubuntu, the ``xserver-xorg-core`` version
+       installed by default is not compatible with Panfrost. But note
+       that upstream Xorg does not work will the blob, so Mesa must be
+       installed so that it is used by default. (see the "Usage"
+       section below). To switch between the upstream and Rockchip
+       versions, run:
+
+.. code-block:: sh
+
+  $ sudo apt install xserver-xorg-core="$(apt-cache show xserver-xorg-core | grep Version | grep -v "$(dpkg -s xserver-xorg-core | grep Version)" | cut -d" " -f2)"
+
+Broken combinations:
+
+#. Panfrost wlroots + Blob Wayland does not work because wlroots does
+   not expose the ``mali_buffer_sharing`` protocol. This might be
+   fixable.
+#. Blob Wayland compositor + Panfrost X11 does not work because the
+   blob does not expose the required protocols for Xwayland
+   acceleration to work
 
 Source
 ------
 
-This repository lives at https://gitlab.freedesktop.org/mesa/mesa.
-Other repositories are likely forks, and code found there is not supported.
+This repository lives at https://gitlab.com/panfork/mesa, and is a
+fork, so not supported by upstream.
 
+Upstream source is at https://gitlab.freedesktop.org/mesa/mesa.
 
-Build & install
----------------
+Depdendencies
+-------------
 
-You can find more information in our documentation (`docs/install.rst
-<https://mesa3d.org/install.html>`_), but the recommended way is to use
-Meson (`docs/meson.rst <https://mesa3d.org/meson.html>`_):
+For Debian-based distributions:
 
 .. code-block:: sh
 
-  $ mkdir build
-  $ cd build
-  $ meson ..
+  $ sudo apt install build-essential meson git python3-mako libexpat1-dev bison flex libwayland-egl-backend-dev libxext-dev libxfixes-dev libxcb-glx0-dev libxcb-shm0-dev libxcb-dri2-0-dev libxcb-dri3-dev libxcb-present-dev libxshmfence-dev libxxf86vm-dev libxrandr-dev
+
+Also needed is ``libdrm`` and ``wayland-protocols``, but those
+packages are too old in Debian Bullseye, and must be compiled from
+source:
+
+.. code-block:: sh
+
+  $ git clone https://gitlab.freedesktop.org/mesa/drm
+  $ mkdir drm/build
+  $ cd drm/build
+  $ meson
+  $ sudo ninja install
+
+.. code-block:: sh
+
+  $ git clone https://gitlab.freedesktop.org/wayland/wayland-protocols
+  $ mkdir wayland-protocols/build
+  $ cd wayland-protocols/build
+  $ git checkout 1.24
+  $ meson
   $ sudo ninja install
 
+Build & install
+---------------
 
-Support
--------
+To install to ``/opt/panfrost``:
 
-Many Mesa devs hang on IRC; if you're not sure which channel is
-appropriate, you should ask your question on `OFTC's #dri-devel
-<irc://irc.oftc.net/dri-devel>`_, someone will redirect you if
-necessary.
-Remember that not everyone is in the same timezone as you, so it might
-take a while before someone qualified sees your question.
-To figure out who you're talking to, or which nick to ping for your
-question, check out `Who's Who on IRC
-<https://dri.freedesktop.org/wiki/WhosWho/>`_.
+.. code-block:: sh
 
-The next best option is to ask your question in an email to the
-mailing lists: `mesa-dev\@lists.freedesktop.org
-<https://lists.freedesktop.org/mailman/listinfo/mesa-dev>`_
+  $ mkdir build
+  $ cd build
+  $ meson -Dgallium-drivers=panfrost -Dvulkan-drivers= -Dllvm=disabled --prefix=/opt/panfrost
+  $ sudo ninja install
 
+Usage
+-----
 
-Bug reports
------------
+To run an application with Panfrost (note the windowing system support
+section above):
 
-If you think something isn't working properly, please file a bug report
-(`docs/bugs.rst <https://mesa3d.org/bugs.html>`_).
+.. code-block:: sh
 
+  $ LD_LIBRARY_PATH=/opt/panfrost/lib/aarch64-linux-gnu glmark2-es2-wayland
 
-Contributing
-------------
+To use Panfrost by default, add the directory where you installed it
+to the library search path:
 
-Contributions are welcome, and step-by-step instructions can be found in our
-documentation (`docs/submittingpatches.rst
-<https://mesa3d.org/submittingpatches.html>`_).
+.. code-block:: sh
 
-Note that Mesa uses gitlab for patches submission, review and discussions.
+  $ echo /opt/panfrost/lib/aarch64-linux-gnu | sudo tee /etc/ld.so.conf.d/0-panfrost.conf
+  $ sudo ldconfig

include/drm-uapi/drm_fourcc.h

@@ -1221,6 +1221,13 @@ drm_fourcc_canonicalize_nvidia_format_mod(__u64 modifier)
  */
 #define AFBC_FORMAT_MOD_USM	(1ULL << 12)
 
+/* AFBC native swizzle
+ *
+ * Indicates that the buffer is using RGBA component order regardless of the
+ * actual format.
+ */
+#define AFBC_FORMAT_MOD_NATIVE_SWIZZLE	(1ULL << 32)
+
 /*
  * Arm Fixed-Rate Compression (AFRC) modifiers
  *

src/compiler/glsl/standalone_scaffolding.cpp

@@ -279,6 +279,9 @@ void initialize_context_to_defaults(struct gl_context *ctx, gl_api api)
    ctx->Const.Program[MESA_SHADER_COMPUTE].MaxUniformComponents = 1024;
    ctx->Const.Program[MESA_SHADER_COMPUTE].MaxInputComponents = 0; /* not used */
    ctx->Const.Program[MESA_SHADER_COMPUTE].MaxOutputComponents = 0; /* not used */
+   ctx->Const.Program[MESA_SHADER_COMPUTE].MaxAtomicBuffers = 16;
+   ctx->Const.Program[MESA_SHADER_COMPUTE].MaxAtomicCounters = 16;
+   ctx->Const.Program[MESA_SHADER_COMPUTE].MaxShaderStorageBlocks = 16;
 
    /* Set up default shader compiler options. */
    struct gl_shader_compiler_options options;

src/drm-shim/device.c

@@ -296,6 +296,10 @@ drm_shim_ioctl(int fd, unsigned long request, void *arg)
    ASSERTED int type = _IOC_TYPE(request);
    int nr = _IOC_NR(request);
 
+   /* Used by kbase; do not claim to be a kbase FD */
+   if (type == 0x80)
+      return -EINVAL;
+
    assert(type == DRM_IOCTL_BASE);
 
    if (nr >= DRM_COMMAND_BASE && nr < DRM_COMMAND_END) {

src/egl/drivers/dri2/egl_dri2.c

@@ -3322,6 +3322,8 @@ dri2_query_wayland_buffer_wl(_EGLDisplay *disp,
    const struct wl_drm_components_descriptor *format;
 
    buffer = wayland_drm_buffer_get(dri2_dpy->wl_server_drm, buffer_resource);
+   if (!buffer)
+           buffer = wayland_drm_buffer_get(dri2_dpy->wl_server_mali, buffer_resource);
    if (!buffer)
       return EGL_FALSE;
 

src/egl/drivers/dri2/egl_dri2.h

@@ -296,8 +296,11 @@ struct dri2_egl_display {
    struct wl_display *wl_dpy_wrapper;
    struct wl_registry *wl_registry;
    struct wl_drm *wl_server_drm;
+   struct wl_drm *wl_server_mali;
    struct wl_drm *wl_drm;
+   struct wl_drm *wl_mali;
    uint32_t wl_drm_version, wl_drm_name;
+   uint32_t wl_mali_version, wl_mali_name;
    struct wl_shm *wl_shm;
    struct wl_event_queue *wl_queue;
    struct zwp_linux_dmabuf_v1 *wl_dmabuf;
@@ -345,6 +348,7 @@ struct dri2_egl_surface {
    struct wl_surface *wl_surface_wrapper;
    struct wl_display *wl_dpy_wrapper;
    struct wl_drm *wl_drm_wrapper;
+   struct wl_drm *wl_mali_wrapper;
    struct wl_callback *throttle_callback;
    struct zwp_linux_dmabuf_feedback_v1 *wl_dmabuf_feedback;
    struct dmabuf_feedback dmabuf_feedback, pending_dmabuf_feedback;

src/egl/drivers/dri2/platform_wayland.c

@@ -748,7 +748,7 @@ dri2_wl_create_window_surface(_EGLDisplay *disp, _EGLConfig *conf,
                                                dri2_surf->base.PresentOpaque);
    assert(visual_idx != -1);
 
-   if (dri2_dpy->wl_dmabuf || dri2_dpy->wl_drm) {
+   if (dri2_dpy->wl_dmabuf || dri2_dpy->wl_drm || dri2_dpy->wl_mali) {
       dri2_surf->format = dri2_wl_visuals[visual_idx].wl_drm_format;
    } else {
       assert(dri2_dpy->wl_shm);
@@ -771,6 +771,16 @@ dri2_wl_create_window_surface(_EGLDisplay *disp, _EGLConfig *conf,
                          dri2_surf->wl_queue);
    }
 
+   if (dri2_dpy->wl_mali) {
+      dri2_surf->wl_mali_wrapper = wl_proxy_create_wrapper(dri2_dpy->wl_mali);
+      if (!dri2_surf->wl_mali_wrapper) {
+         _eglError(EGL_BAD_ALLOC, "dri2_create_surface");
+         goto cleanup_queue;
+      }
+      wl_proxy_set_queue((struct wl_proxy *)dri2_surf->wl_mali_wrapper,
+                         dri2_surf->wl_queue);
+   }
+
    dri2_surf->wl_dpy_wrapper = wl_proxy_create_wrapper(dri2_dpy->wl_dpy);
    if (!dri2_surf->wl_dpy_wrapper) {
       _eglError(EGL_BAD_ALLOC, "dri2_create_surface");
@@ -907,6 +917,8 @@ dri2_wl_destroy_surface(_EGLDisplay *disp, _EGLSurface *surf)
    wl_proxy_wrapper_destroy(dri2_surf->wl_dpy_wrapper);
    if (dri2_surf->wl_drm_wrapper)
       wl_proxy_wrapper_destroy(dri2_surf->wl_drm_wrapper);
+   if (dri2_surf->wl_mali_wrapper)
+      wl_proxy_wrapper_destroy(dri2_surf->wl_mali_wrapper);
    if (dri2_surf->wl_dmabuf_feedback) {
       zwp_linux_dmabuf_feedback_v1_destroy(dri2_surf->wl_dmabuf_feedback);
       dmabuf_feedback_fini(&dri2_surf->dmabuf_feedback);
@@ -1503,6 +1515,26 @@ create_wl_buffer(struct dri2_egl_display *dri2_dpy,
       ret = zwp_linux_buffer_params_v1_create_immed(params, width, height,
                                                     fourcc, 0);
       zwp_linux_buffer_params_v1_destroy(params);
+   } else if (dri2_surf->wl_mali_wrapper || dri2_dpy->wl_mali) {
+      struct wl_drm *wl_mali =
+         dri2_surf ? dri2_surf->wl_mali_wrapper : dri2_dpy->wl_mali;
+      int fd = -1, stride;
+
+      if (num_planes > 1)
+         return NULL;
+
+      query = dri2_dpy->image->queryImage(image, __DRI_IMAGE_ATTRIB_FD, &fd);
+      query &= dri2_dpy->image->queryImage(image, __DRI_IMAGE_ATTRIB_STRIDE, &stride);
+      if (!query) {
+         if (fd >= 0)
+            close(fd);
+         return NULL;
+      }
+
+      ret = mali_buffer_sharing_create_buffer((void *)wl_mali,
+                                              width, height, stride,
+                                              fourcc, 0, 0, fd);
+      close(fd);
    } else {
       struct wl_drm *wl_drm =
          dri2_surf ? dri2_surf->wl_drm_wrapper : dri2_dpy->wl_drm;
@@ -1776,6 +1808,62 @@ drm_handle_device(void *data, struct wl_drm *drm, const char *device)
    }
 }
 
+static void
+mali_handle_device(void *data, struct mali_buffer_sharing *drm, const char *device)
+{
+   struct dri2_egl_display *dri2_dpy = data;
+   drm_magic_t magic;
+
+   // hack
+   //printf("device '%s'\n", device);
+   dri2_dpy->device_name = strdup("/dev/dri/card0");
+
+   dri2_dpy->fd = loader_open_device(dri2_dpy->device_name);
+   if (dri2_dpy->fd == -1) {
+      _eglLog(_EGL_WARNING, "wayland-egl: could not open %s (%s)",
+              dri2_dpy->device_name, strerror(errno));
+      free(dri2_dpy->device_name);
+      dri2_dpy->device_name = NULL;
+      return;
+   }
+
+   if (drmGetNodeTypeFromFd(dri2_dpy->fd) == DRM_NODE_RENDER) {
+      dri2_dpy->authenticated = true;
+   } else {
+      roundtrip(dri2_dpy);
+      if (drmGetMagic(dri2_dpy->fd, &magic)) {
+         close(dri2_dpy->fd);
+         dri2_dpy->fd = -1;
+         free(dri2_dpy->device_name);
+         dri2_dpy->device_name = NULL;
+         _eglLog(_EGL_WARNING, "wayland-egl: drmGetMagic failed");
+         return;
+      }
+
+      mali_buffer_sharing_auth((void *)dri2_dpy->wl_mali, magic);
+      roundtrip(dri2_dpy);
+      // We don't get a callback
+      dri2_dpy->authenticated = true;
+   }
+
+   int supported_fourcc[] = {
+      WL_DRM_FORMAT_ABGR16F,
+      WL_DRM_FORMAT_ABGR2101010,
+      WL_DRM_FORMAT_XRGB8888,
+      WL_DRM_FORMAT_ARGB8888,
+      WL_DRM_FORMAT_ABGR8888,
+      WL_DRM_FORMAT_XBGR8888,
+      WL_DRM_FORMAT_RGB565,
+   };
+
+   for (unsigned i = 0; i < ARRAY_SIZE(supported_fourcc); ++i) {
+      int visual_idx = dri2_wl_visual_idx_from_fourcc(supported_fourcc[i]);
+      assert(visual_idx != -1);
+
+      BITSET_SET(dri2_dpy->formats.formats_bitmap, visual_idx);
+   }
+}
+
 static void
 drm_handle_format(void *data, struct wl_drm *drm, uint32_t format)
 {
@@ -2836,6 +2924,8 @@ dri2_teardown_wayland(struct dri2_egl_display *dri2_dpy)
    dri2_wl_formats_fini(&dri2_dpy->formats);
    if (dri2_dpy->wl_drm)
       wl_drm_destroy(dri2_dpy->wl_drm);
+   if (dri2_dpy->wl_mali)
+      wl_drm_destroy(dri2_dpy->wl_mali);
    if (dri2_dpy->wl_dmabuf)
       zwp_linux_dmabuf_v1_destroy(dri2_dpy->wl_dmabuf);
    if (dri2_dpy->wl_shm)

src/gallium/auxiliary/cso_cache/cso_context.c

@@ -1446,6 +1446,11 @@ cso_single_sampler(struct cso_context *cso, enum pipe_shader_type shader_stage,
    }
 }
 
+void
+cso_set_max_sampler(struct cso_context *ctx, int max_sampler_seen)
+{
+   ctx->max_sampler_seen = max_sampler_seen;
+}
 
 /**
  * Send staged sampler state to the driver.

src/gallium/auxiliary/cso_cache/cso_context.h

@@ -84,6 +84,9 @@ void
 cso_single_sampler(struct cso_context *cso, enum pipe_shader_type shader_stage,
                    unsigned idx, const struct pipe_sampler_state *states);
 
+void
+cso_set_max_sampler(struct cso_context *ctx, int max_sampler_seen);
+
 void
 cso_single_sampler_done(struct cso_context *cso,
                         enum pipe_shader_type shader_stage);