dma-buf new API consists of: - mmap(dma_buf_fd, ...): the ability to map a dma-buf file-descriptor of a graphics buffer to the userspace, and more importantly, to actually write on the mapped pointer (which was not possible before). It’s worth noting that the Direct Rendering Manager (DRM) and the hardware driver implementation are fundamentally important to safely export the graphics handle to be mapped. - ioctl(dma_buf_fd, DMA_BUF_IOCTL_SYNC, &args): cache coherency management in cases where the CPU and GPU devices are being accessed through dma-buf at the same time. Coherency markers, which forward directly to existing dma-buf device drivers vfunc hooks, are exposed to the userspace through the DMA_BUF_IOCTL_SYNC ioctl and have to be used before and after the mapped area is accessed. This is fundamentally important in hardware architectures where the graphics engine and the CPU cores don't share caches but also important in other type of hardware where the memory hierarchy is (most of the time) coherent. More details can be found in this patch set: http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=c11e391da2a8fe973c3c2398452000bed505851e v2: use uint32_t for color type, increment the variable and add --interactive-debug=paint v3: use igt_display_commit() to mode set the crtc so the rectangle is shown painted; also added Testcase description on the beginning of the file. v4: remove crtc actually which seems superfluous; add a igt_skip_on in case support for dma-buf mmap is nonexistent. Signed-off-by: Tiago Vignatti <tiago.vignatti@xxxxxxxxx> --- tests/Makefile.sources | 1 + tests/prime_mmap_kms.c | 275 +++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 276 insertions(+) create mode 100644 tests/prime_mmap_kms.c diff --git a/tests/Makefile.sources b/tests/Makefile.sources index 9ea9b5b..7b5f316 100644 --- a/tests/Makefile.sources +++ b/tests/Makefile.sources @@ -113,6 +113,7 @@ TESTS_progs_M = \ pm_sseu \ prime_mmap \ prime_mmap_coherency \ + prime_mmap_kms \ prime_self_import \ template \ $(NULL) diff --git a/tests/prime_mmap_kms.c b/tests/prime_mmap_kms.c new file mode 100644 index 0000000..8e14b25 --- /dev/null +++ b/tests/prime_mmap_kms.c @@ -0,0 +1,275 @@ +/* + * Copyright © 2016 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + * + * Authors: + * Tiago Vignatti <tiago.vignatti at intel.com> + */ + +/* + * Testcase: show case dma-buf new API and processes restrictions. Most likely + * you want to run like ./prime_mmap_kms --interactive-debug=paint, to see the + * actual rectangle painted on the screen. + */ + +#include "igt.h" + +IGT_TEST_DESCRIPTION( + "Efficiently sharing CPU and GPU buffers"); + +/* + * render_process_t: + * + * Render is basically a user-space regular client. It's the unprivileged + * process with limited system accesses. + * + * Worth note the vendor-independent characteristic, meaning that the + * client doesn't need to perform any vendor specific calls for buffer + * handling. Mesa GBM library is a counter-example because, even though its API + * is vendor-independent, under-the-hood the library actually calls vendor + * specific ioctls, which is not really sandboxable and not the goal here. + */ +typedef struct { + int prime_fd; + size_t size; + int width; + int height; +} render_process_t; + +typedef struct { + int x; + int y; + int w; + int h; +} rect_t; + +/* set ptr in a linear view */ +static void set_pixel(void *_ptr, int index, uint32_t color, int bpp) +{ + if (bpp == 16) { + uint16_t *ptr = _ptr; + ptr[index] = color; + } else if (bpp == 32) { + uint32_t *ptr = _ptr; + ptr[index] = color; + } else { + igt_assert_f(false, "bpp: %d\n", bpp); + } +} + +static void paint(render_process_t *render) +{ + void *frame; + rect_t rect = { + .x = 200, + .y = 200, + .w = render->width / 4, + .h = render->height / 4, + }; + uint32_t color = 0xFF; + int stride, bpp; + int x, y, line_begin; + + frame = mmap(NULL, render->size, PROT_READ | PROT_WRITE, MAP_SHARED, + render->prime_fd, 0); + igt_assert(frame != MAP_FAILED); + + /* TODO: what's the mmap'ed buffer semantics on tiling, format etc. How + * does the client know whether that the BO was created X-tiled, + * Y-tiled and how it will map back? This is something we need to + * address in this API still. */ + stride = render->width * 4; + bpp = 32; + + /* ioctls to keep up the GPU <-> CPU coherency */ + prime_sync_start(render->prime_fd, true); + + /* the actual painting phase happens here */ + for (y = rect.y; y < rect.y + rect.h; y++) { + line_begin = y * stride / (bpp / 8); + for (x = rect.x; x < rect.x + rect.w; x++) + set_pixel(frame, line_begin + x, color, bpp); + } + + prime_sync_end(render->prime_fd, true); + munmap(frame, render->size); +} + +static void init_renderer(int prime_fd, int fb_size, int width, int height) +{ + render_process_t render; + + render.prime_fd = prime_fd; + render.size = fb_size; + render.width = width; + render.height = height; + paint(&render); +} + +/* + * gpu_process_t: + * + * GPU process is the privileged process and has access to the system graphics + * routines, like DRM, display management and driver accesses. + */ +typedef struct { + int drm_fd; + igt_display_t display; + struct igt_fb fb; + igt_output_t *output; + igt_plane_t *primary; + enum pipe pipe; +} gpu_process_t; + +static void cleanup_crtc(gpu_process_t *gpu) +{ + igt_display_t *display = &gpu->display; + igt_output_t *output = gpu->output; + + igt_plane_set_fb(gpu->primary, NULL); + + igt_output_set_pipe(output, PIPE_ANY); + igt_display_commit(display); + + igt_remove_fb(gpu->drm_fd, &gpu->fb); +} + +static bool prepare_crtc(gpu_process_t *gpu) +{ + igt_display_t *display = &gpu->display; + igt_output_t *output = gpu->output; + drmModeModeInfo *mode; + + /* select the pipe we want to use */ + igt_output_set_pipe(output, gpu->pipe); + igt_display_commit(display); + + if (!output->valid) { + igt_output_set_pipe(output, PIPE_ANY); + igt_display_commit(display); + return false; + } + + mode = igt_output_get_mode(output); + + /* create a white fb and flip to it */ + igt_create_color_fb(gpu->drm_fd, mode->hdisplay, mode->vdisplay, + DRM_FORMAT_XRGB8888, LOCAL_DRM_FORMAT_MOD_NONE, + 1.0, 1.0, 1.0, &gpu->fb); + + gpu->primary = igt_output_get_plane(output, IGT_PLANE_PRIMARY); + + igt_plane_set_fb(gpu->primary, &gpu->fb); + igt_display_commit(display); + + return true; +} + +/* + * The idea is to create a BO (in this case the framebuffer's) in one process, + * export and pass its prime fd to another process, which in turn uses the fd + * to map and write. This is Chrome-like architectures, where the Web content + * (a "tab" or the "unprivileged process") maps and CPU-paints a buffer, which + * was previously allocated in the GPU process ("privileged process"). + */ +static void run_test(gpu_process_t *gpu) +{ + igt_display_t *display = &gpu->display; + igt_output_t *output; + enum pipe pipe; + int prime_fd; + + for_each_connected_output(display, output) { + gpu->output = output; + for_each_pipe(display, pipe) { + gpu->pipe = pipe; + + if (!prepare_crtc(gpu)) + continue; + + prime_fd = prime_handle_to_fd_for_mmap(gpu->drm_fd, + gpu->fb.gem_handle); + igt_skip_on(prime_fd == -1 && errno == EINVAL); + + /* Note that it only shares the dma-buf fd and some + * other basic info */ + igt_fork(renderer_no, 1) { + init_renderer(prime_fd, gpu->fb.size, gpu->fb.width, + gpu->fb.height); + } + igt_waitchildren(); + + igt_debug_wait_for_keypress("paint"); + cleanup_crtc(gpu); + + /* once is enough */ + return; + } + } + + igt_skip("no valid crtc/connector combinations found\n"); +} + +static int +check_for_dma_buf_mmap(int fd) +{ + int dma_buf_fd; + char *ptr; + uint32_t handle; + int ret = 1; + + handle = gem_create(fd, 4096); + dma_buf_fd = prime_handle_to_fd(fd, handle); + ptr = mmap(NULL, 4096, PROT_READ, MAP_SHARED, dma_buf_fd, 0); + if (ptr != MAP_FAILED) + ret = 0; + munmap(ptr, 4096); + gem_close(fd, handle); + close(dma_buf_fd); + return ret; +} + +igt_main +{ + gpu_process_t gpu; + + igt_skip_on_simulation(); + + igt_fixture { + gpu.drm_fd = drm_open_driver_master(DRIVER_INTEL); + igt_skip_on((check_for_dma_buf_mmap(gpu.drm_fd) != 0)); + kmstest_set_vt_graphics_mode(); + + igt_require_pipe_crc(); + + igt_display_init(&gpu.display, gpu.drm_fd); + } + + igt_subtest("buffer-sharing") + run_test(&gpu); + + igt_fixture { + igt_display_fini(&gpu.display); + close(gpu.drm_fd); + } + + igt_exit(); +} -- 2.7.4 _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
