tests/gem_ppgtt: New gem_ppgtt set of tests

This test covers some tricky corner cases found during
the True-and-only PPGTT feature development.

v2: Add pthread requirement to makefile.
v3: Added new "pinned" testcase.
v4 (Ben): Update rendercpy to use contexts

Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>

4 files changed, 327 insertions, 0 deletions

diff --git a/tests/.gitignore b/tests/.gitignore
index 4448cb64..941760d2 100644
--- a/tests/.gitignore
+++ b/tests/.gitignore
@@ -54,6 +54,7 @@ gem_partial_pwrite_pread
 gem_persistent_relocs
 gem_pin
 gem_pipe_control_store_loop
+gem_ppgtt
 gem_pread
 gem_pread_after_blit
 gem_pwrite
diff --git a/tests/Makefile.am b/tests/Makefile.am
index b8cddd56..287391e1 100644
--- a/tests/Makefile.am
+++ b/tests/Makefile.am
@@ -32,6 +32,7 @@ LDADD = ../lib/libintel_tools.la $(PCIACCESS_LIBS) $(DRM_LIBS)
 LDADD += $(CAIRO_LIBS) $(LIBUDEV_LIBS) $(GLIB_LIBS)
 AM_CFLAGS += $(CAIRO_CFLAGS) $(LIBUDEV_CFLAGS) $(GLIB_CFLAGS)
 
+gem_ppgtt_LDADD = $(LDADD) -lpthread
 gem_close_race_LDADD = $(LDADD) -lpthread
 gem_fence_thrash_CFLAGS = $(AM_CFLAGS) $(THREAD_CFLAGS)
 gem_fence_thrash_LDADD = $(LDADD) -lpthread
diff --git a/tests/Makefile.sources b/tests/Makefile.sources
index aa34aeb2..3eb50dd3 100644
--- a/tests/Makefile.sources
+++ b/tests/Makefile.sources
@@ -36,6 +36,7 @@ TESTS_progs_M = \
 	gem_mmap_gtt \
 	gem_partial_pwrite_pread \
 	gem_persistent_relocs \
+	gem_ppgtt \
 	gem_pread \
 	gem_pread_after_blit \
 	gem_pwrite \
diff --git a/tests/gem_ppgtt.c b/tests/gem_ppgtt.c
new file mode 100644
index 00000000..266a6a32
--- /dev/null
+++ b/tests/gem_ppgtt.c
@@ -0,0 +1,324 @@
+/*
+ * Copyright © 2013 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:
+ *    Oscar Mateo <oscar.mateo@intel.com>
+ *
+ */
+
+/*
+ * This test covers some tricky corner cases found during
+ * True PPGTT development
+ */
+
+#include <pthread.h>
+#include "rendercopy.h"
+#include "igt_debugfs.h"
+
+#define WIDTH 512
+#define STRIDE (WIDTH*4)
+#define HEIGHT 512
+#define SIZE (HEIGHT*STRIDE)
+
+/* options */
+int option_num_threads = 10;
+int option_num_iter = 200;
+bool option_tiled = false;
+
+/* globals */
+int global_fd;
+int devid;
+render_copyfunc_t rendercopy;
+
+static void init_buffer(drm_intel_bufmgr *bufmgr,
+			struct scratch_buf *buf, bool special)
+{
+	uint32_t tiling;
+	unsigned long pitch = STRIDE;
+
+	if (option_tiled)
+	{
+		tiling = I915_TILING_X;
+		buf->bo = drm_intel_bo_alloc_tiled(bufmgr, "",
+			     WIDTH, HEIGHT, 4,
+			     &tiling, &pitch, 0);
+	}
+	else
+	{
+		tiling = I915_TILING_NONE;
+		buf->bo = drm_intel_bo_alloc(bufmgr, "",
+				special? 16*SIZE : SIZE, 4096);
+	}
+
+	igt_assert(buf->bo);
+	buf->size = SIZE;
+	buf->tiling = tiling;
+	buf->stride = pitch;
+}
+
+static bool check_buffer_contents(struct scratch_buf *buf, uint32_t val)
+{
+	uint32_t *ptr;
+	int i;
+	drm_intel_bo *bo = buf->bo;
+
+	if (buf->tiling == I915_TILING_NONE)
+		do_or_die(drm_intel_bo_map(bo, 0));
+	else
+		do_or_die(drm_intel_gem_bo_map_gtt(bo));
+
+	ptr = bo->virtual;
+	for (i = 0; i < WIDTH*HEIGHT; i++) {
+		if (ptr[i] != val) {
+			printf("Expected 0x%08x, found 0x%08x at offset 0x%08x\n",
+					val, ptr[i], i * 4);
+			return false;
+		}
+		val++;
+	}
+
+	if (buf->tiling == I915_TILING_NONE)
+		drm_intel_bo_unmap(bo);
+	else
+		drm_intel_gem_bo_unmap_gtt(bo);
+
+	return true;
+}
+
+static void init_buffer_contents(struct scratch_buf *buf, uint32_t val)
+{
+	uint32_t *ptr;
+	int i;
+	drm_intel_bo *bo = buf->bo;
+
+	if (buf->tiling == I915_TILING_NONE)
+		do_or_die(drm_intel_bo_map(bo, 1));
+	else
+		do_or_die(drm_intel_gem_bo_map_gtt(bo));
+
+	ptr = bo->virtual;
+	for (i = 0; i < WIDTH*HEIGHT; i++)
+		ptr[i] = val++;
+
+	if (buf->tiling == I915_TILING_NONE)
+		drm_intel_bo_unmap(bo);
+	else
+		drm_intel_gem_bo_unmap_gtt(bo);
+}
+
+static void run(void)
+{
+	int fd;
+	drm_intel_bufmgr *bufmgr;
+	struct intel_batchbuffer *batch;
+	struct scratch_buf src, dst;
+	int i;
+	uint32_t start = 0;
+
+	/* Initialize */
+	igt_assert((fd = drm_open_any()) >= 0);
+	bufmgr = drm_intel_bufmgr_gem_init(fd, 4096);
+	batch = intel_batchbuffer_alloc(bufmgr, devid);
+	init_buffer(bufmgr, &src, false);
+	init_buffer_contents(&src, start);
+	init_buffer(bufmgr, &dst, false);
+
+	/* Work */
+	for (i = 0; i < option_num_iter; i++)
+		rendercopy(batch, NULL, &src, 0, 0, WIDTH, HEIGHT, &dst, 0, 0);
+	igt_assert(check_buffer_contents(&dst, start));
+
+	/* Cleanup */
+	drm_intel_bo_unreference(src.bo);
+	drm_intel_bo_unreference(dst.bo);
+	intel_batchbuffer_free(batch);
+	drm_intel_bufmgr_destroy(bufmgr);
+	close(fd);
+}
+
+static void *work_multiple(void *arg)
+{
+	run();
+
+	pthread_exit(NULL);
+}
+
+static void run_multiple(void)
+{
+	int i = 0;
+	void *retval;
+	pthread_t *threads = calloc(option_num_threads, sizeof(*threads));
+
+	for (i = 0; i < option_num_threads; i++)
+		pthread_create(&threads[i], NULL, work_multiple, &i);
+
+	for (i = 0; i < option_num_threads; i++)
+		igt_assert(pthread_join(threads[i], &retval) == 0);
+
+	free(threads);
+}
+
+static void run_with_namedbo(uint32_t name)
+{
+	int fd;
+	drm_intel_bufmgr *bufmgr;
+	struct intel_batchbuffer *batch;
+	struct scratch_buf src, dst;
+	uint32_t start = 0;
+	int i;
+
+	/* Initialize */
+	igt_assert((fd = drm_open_any()) >= 0);
+	bufmgr = drm_intel_bufmgr_gem_init(fd, 4096);
+	batch = intel_batchbuffer_alloc(bufmgr, devid);
+	init_buffer(bufmgr, &dst, false);
+
+	/* Recover src */
+	src.size = dst.size;
+	src.tiling = dst.tiling;
+	src.stride = dst.stride;
+	src.bo = drm_intel_bo_gem_create_from_name(bufmgr, "", name);
+	igt_assert(src.bo);
+	igt_assert(check_buffer_contents(&src, start));
+
+	/* Work */
+	for (i = 0; i < option_num_iter; i++)
+		rendercopy(batch, NULL, &src, 0, 0, WIDTH, HEIGHT, &dst, 0, 0);
+	igt_assert(check_buffer_contents(&dst, start));
+
+	/* Cleanup */
+	drm_intel_bo_unreference(src.bo);
+	drm_intel_bo_unreference(dst.bo);
+	intel_batchbuffer_free(batch);
+	drm_intel_bufmgr_destroy(bufmgr);
+	close(fd);
+}
+
+static void *work_flinked(void *arg)
+{
+	uint32_t *name = (uint32_t *)arg;
+
+	run_with_namedbo(*name);
+
+	pthread_exit(NULL);
+}
+
+static void run_flinked(void)
+{
+	drm_intel_bufmgr *bufmgr;
+	struct scratch_buf src;
+	uint32_t name;
+	uint32_t start = 0;
+	int i;
+	void *retval;
+	pthread_t *threads = calloc(option_num_threads, sizeof(*threads));
+
+	bufmgr = drm_intel_bufmgr_gem_init(global_fd, 4096);
+	init_buffer(bufmgr, &src, false);
+	init_buffer_contents(&src, start);
+	do_or_die(drm_intel_bo_flink(src.bo, &name));
+
+	for (i = 0; i < option_num_threads; i++)
+		pthread_create(&threads[i], NULL, work_flinked, &name);
+
+	for (i = 0; i < option_num_threads; i++)
+		igt_assert(pthread_join(threads[i], &retval) == 0);
+
+	free(threads);
+	drm_intel_bo_unreference(src.bo);
+	drm_intel_bufmgr_destroy(bufmgr);
+}
+
+static void run_pinned(void)
+{
+	drm_intel_bufmgr *bufmgr;
+	struct scratch_buf src;
+	uint32_t name;
+	uint32_t val = 0;
+
+	bufmgr = drm_intel_bufmgr_gem_init(global_fd, 4096);
+	init_buffer(bufmgr, &src, true);
+	init_buffer_contents(&src, val);
+	do_or_die(drm_intel_bo_flink(src.bo, &name));
+
+	/* Pin to GGTT */
+	drm_intel_bo_pin(src.bo, 4096);
+
+	/* Use the source bo in a different context (pin to a PPGTT) */
+	run_with_namedbo(name);
+
+	/* Make sure we switch the context in every ring so that the
+	 * previous context finally dies */
+	gem_quiescent_gpu(global_fd);
+	igt_drop_caches_set(DROP_RETIRE);
+
+	drm_intel_bo_unpin(src.bo);	
+	drm_intel_bo_unreference(src.bo);
+	drm_intel_bufmgr_destroy(bufmgr);
+}
+
+static void do_tests(bool use_tiled, const char *suffix)
+{
+	option_tiled = use_tiled;
+
+	/* One context, rendercopying in a loop */
+	igt_subtest_f("single-ctx-%s", suffix)
+		run();
+
+	/* Multiple threads, each with its own context
+	 * rendercopying in a loop */
+	igt_subtest_f("multiple-ctx-%s", suffix) {
+		run_multiple();
+	}
+
+	/* Multiple threads, each with its own context
+	 * rendercopying in a loop. The src bo is shared
+	 * among them */
+	igt_subtest_f("multiple-ctx-flink-%s", suffix) {
+		run_flinked();
+	}
+
+	/* Pin object to GGTT, create a context, bind the
+	 * object to its PPGTT and then clean the context */
+	igt_subtest_f("single-ctx-pinned-%s", suffix) {
+		run_pinned();
+	}
+}
+
+igt_main
+{
+	igt_fixture {
+		global_fd = drm_open_any();
+		devid = intel_get_drm_devid(global_fd);
+		rendercopy = get_render_copyfunc(devid);
+		igt_require(rendercopy);
+		if (igt_run_in_simulation())
+			option_num_iter = 10;
+			option_num_threads = 3;
+	}
+
+	do_tests(false, "linear");
+	do_tests(true, "tiled");
+
+	igt_fixture
+		close(global_fd);
+}