Add a test for pread after blitting to an object.

This caught a bug with get_user_pages usage in the kernel, which would
result in zeroes being read out of the object when faulting in a new page.

.gitignore

@@ -38,6 +38,7 @@ tests/getversion
 tests/gem_basic
 tests/gem_flink
 tests/gem_mmap
+tests/gem_pread_after_blit
 tests/gem_readwrite
 tests/gem_tiled_blits
 tools/intel_gpu_dump

lib/intel_gpu_tools.c

@@ -32,8 +32,9 @@
 #include <assert.h>
 #include <sys/ioctl.h>
 #include "intel_gpu_tools.h"
-#include "intel_chipset.h"
 #include "i915_drm.h"
+#include "intel_batchbuffer.h"
+#include "intel_chipset.h"
 
 struct pci_device *pci_dev;
 uint32_t devid;
@@ -98,3 +99,48 @@ intel_get_mmio(void)
 		exit(1);
 	}
 }
+
+void
+intel_copy_bo(struct intel_batchbuffer *batch,
+	      drm_intel_bo *dst_bo, drm_intel_bo *src_bo,
+	      int width, int height)
+{
+	uint32_t src_tiling, dst_tiling, swizzle;
+	uint32_t src_pitch, dst_pitch;
+	uint32_t cmd_bits = 0;
+
+	drm_intel_bo_get_tiling(src_bo, &src_tiling, &swizzle);
+	drm_intel_bo_get_tiling(dst_bo, &dst_tiling, &swizzle);
+
+	src_pitch = width * 4;
+	if (IS_965(devid) && src_tiling != I915_TILING_NONE) {
+		src_pitch /= 4;
+		cmd_bits |= XY_SRC_COPY_BLT_SRC_TILED;
+	}
+
+	dst_pitch = width * 4;
+	if (IS_965(devid) && dst_tiling != I915_TILING_NONE) {
+		dst_pitch /= 4;
+		cmd_bits |= XY_SRC_COPY_BLT_DST_TILED;
+	}
+
+	BEGIN_BATCH(8);
+	OUT_BATCH(XY_SRC_COPY_BLT_CMD |
+		  XY_SRC_COPY_BLT_WRITE_ALPHA |
+		  XY_SRC_COPY_BLT_WRITE_RGB |
+		  cmd_bits);
+	OUT_BATCH((3 << 24) | /* 32 bits */
+		  (0xcc << 16) | /* copy ROP */
+		  dst_pitch);
+	OUT_BATCH(0); /* dst x1,y1 */
+	OUT_BATCH((height << 16) | width); /* dst x2,y2 */
+	OUT_RELOC(dst_bo, I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER, 0);
+	OUT_BATCH(0); /* src x1,y1 */
+	OUT_BATCH(src_pitch);
+	OUT_RELOC(src_bo, I915_GEM_DOMAIN_RENDER, 0, 0);
+	ADVANCE_BATCH();
+
+	intel_batchbuffer_flush(batch);
+}
+
+

lib/intel_gpu_tools.h

@@ -27,6 +27,8 @@
 
 #include <sys/types.h>
 #include <pciaccess.h>
+#include "i915_drm.h"
+#include "intel_batchbuffer.h"
 #include "intel_chipset.h"
 #include "intel_reg.h"
 
@@ -42,3 +44,6 @@ INREG(uint32_t reg)
 
 void intel_get_mmio(void);
 void intel_get_drm_devid(int fd);
+void intel_copy_bo(struct intel_batchbuffer *batch,
+		   drm_intel_bo *dst_bo, drm_intel_bo *src_bo,
+		   int width, int height);

tests/Makefile.am

@@ -5,6 +5,7 @@ TESTS = getversion \
 	gem_flink \
 	gem_readwrite \
 	gem_mmap \
+	gem_pread_after_blit \
 	gem_tiled_blits
 
 EXTRA_PROGRAMS = $(TESTS)

tests/gem_pread_after_blit.c

@@ -0,0 +1,176 @@
+/*
+ * Copyright © 2009 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:
+ *    Eric Anholt <eric@anholt.net>
+ *
+ */
+
+/** @file gem_pread_after_blit.c
+ *
+ * This is a test of pread's behavior when getting values out of just-drawn-to
+ * buffers.
+ *
+ * The goal is to catch failure in the whole-buffer-flush or
+ * ranged-buffer-flush paths in the kernel.
+ */
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <assert.h>
+#include <fcntl.h>
+#include <inttypes.h>
+#include <errno.h>
+#include <sys/stat.h>
+#include <sys/time.h>
+#include "drm.h"
+#include "i915_drm.h"
+#include "drmtest.h"
+#include "intel_bufmgr.h"
+#include "intel_batchbuffer.h"
+#include "intel_gpu_tools.h"
+
+static drm_intel_bufmgr *bufmgr;
+struct intel_batchbuffer *batch;
+static const int width = 512, height = 512;
+static const int size = 1024 * 1024;
+
+#define PAGE_SIZE 4096
+
+static drm_intel_bo *
+create_bo(uint32_t val)
+{
+	drm_intel_bo *bo;
+	uint32_t *vaddr;
+	int i;
+
+	bo = drm_intel_bo_alloc(bufmgr, "src bo", size, 4096);
+
+	/* Fill the BO with dwords starting at start_val */
+	drm_intel_bo_map(bo, 1);
+	vaddr = bo->virtual;
+
+	for (i = 0; i < 1024 * 1024 / 4; i++)
+		vaddr[i] = val++;
+
+	drm_intel_bo_unmap(bo);
+
+	return bo;
+}
+
+static void
+verify_large_read(drm_intel_bo *bo, uint32_t val)
+{
+	uint32_t buf[size / 4];
+	int i;
+
+	drm_intel_bo_get_subdata(bo, 0, size, buf);
+
+	for (i = 0; i < size / 4; i++) {
+		if (buf[i] != val) {
+			fprintf(stderr,
+				"Unexpected value 0x%08x instead of "
+				"0x%08x at offset 0x%08x (%p)\n",
+				buf[i], val, i * 4, buf);
+			abort();
+		}
+		val++;
+	}
+}
+
+/** This reads at the size that Mesa usees for software fallbacks. */
+static void
+verify_small_read(drm_intel_bo *bo, uint32_t val)
+{
+	uint32_t buf[4096 / 4];
+	int offset, i;
+
+	for (i = 0; i < 4096 / 4; i++)
+		buf[i] = 0x00c0ffee;
+
+	for (offset = 0; offset < size; offset += PAGE_SIZE) {
+		drm_intel_bo_get_subdata(bo, offset, PAGE_SIZE, buf);
+
+		for (i = 0; i < PAGE_SIZE; i += 4) {
+			if (buf[i / 4] != val) {
+				fprintf(stderr,
+					"Unexpected value 0x%08x instead of "
+					"0x%08x at offset 0x%08x\n",
+					buf[i / 4], val, i * 4);
+				abort();
+			}
+			val++;
+		}
+	}
+}
+
+int
+main(int argc, char **argv)
+{
+	int fd;
+	drm_intel_bo *src1, *src2, *bo;
+	uint32_t start1 = 0;
+	uint32_t start2 = 1024 * 1024 / 4;
+
+	fd = drm_open_any();
+	intel_get_drm_devid(fd);
+
+	bufmgr = drm_intel_bufmgr_gem_init(fd, 4096);
+	drm_intel_bufmgr_gem_enable_reuse(bufmgr);
+	batch = intel_batchbuffer_alloc(bufmgr);
+
+	src1 = create_bo(start1);
+	src2 = create_bo(start2);
+
+	bo = drm_intel_bo_alloc(bufmgr, "dst bo", size, 4096);
+
+	/* First, do a full-buffer read after blitting */
+	printf("Large read after blit 1\n");
+	intel_copy_bo(batch, bo, src1, width, height);
+	verify_large_read(bo, start1);
+	printf("Large read after blit 2\n");
+	intel_copy_bo(batch, bo, src2, width, height);
+	verify_large_read(bo, start2);
+
+	printf("Small reads after blit 1\n");
+	intel_copy_bo(batch, bo, src1, width, height);
+	verify_small_read(bo, start1);
+	printf("Small reads after blit 2\n");
+	intel_copy_bo(batch, bo, src2, width, height);
+	verify_small_read(bo, start2);
+
+	printf("Large read after blit 3\n");
+	intel_copy_bo(batch, bo, src1, width, height);
+	verify_large_read(bo, start1);
+
+	drm_intel_bo_unreference(src1);
+	drm_intel_bo_unreference(src2);
+	drm_intel_bo_unreference(bo);
+
+	intel_batchbuffer_free(batch);
+	drm_intel_bufmgr_destroy(bufmgr);
+
+	close(fd);
+
+	return 0;
+}

tools/Makefile.am

@@ -3,8 +3,7 @@ bin_PROGRAMS = \
 	intel_gpu_top \
 	intel_stepping
 
-intel_gpu_top_LDADD = ../lib/libintel_tools.la $(PCIACCESS_LIBS)
-intel_stepping_LDADD = $(PCIACCESS_LIBS)
+LDADD = ../lib/libintel_tools.la $(DRM_LIBS) $(PCIACCESS_LIBS)
 
 AM_CFLAGS = $(DRM_CFLAGS) $(PCIACCESS_CFLAGS) $(WARN_CFLAGS) \
 	-I$(srcdir)/.. \