[PATCH I-g-t V4 0/2] Tests: Add test cases based on multi drm_fd to test sync

[PATCH I-g-t V4 0/2] Tests: Add test cases based on multi drm_fd to test sync

[Zhao Yakui]

This follows Daniel's advice to add the two test cases based on multi drm_fd to 
test the ring sync and CPU<->GPU sync.
The Broadwell GT3 machine has two independent BSD rings that can be used
to process the video commands. This is implemented in kernel driver and transparent
to the user-space. But we still need to check the ring sync and CPU<->GPU sync for
the second BSD ring. Two tests are created based on the multi drm_fds to
test the sync. Multi drm_fd can assure that the second BSD ring has the opportunity
to dispatch the GPU command. 

V1->V2: Follow Daniel's comment to add one subtext instead of one individual
test case, which is used to test the CPU<->GPU sync under multi BSD rings/

V2->V3: Follow Imre's comment to remove the unnecessary initialization and
use igt_assert_f instead of igt_assert

V3->V4: Add gem_multi_bsd_sync_loop.c into the tests/.gitignore

Zhao Yakui (2):
  tests: Add one ring sync case based on multi drm_fd to test ring
    semaphore sync under multi BSD rings
  tests/gem_dummy_reloc_loop: Add one subtest based on multi drm_fd to
    test CPU<->GPU sync under multi BSD rings

 tests/.gitignore                |    1 +
 tests/Makefile.sources          |    1 +
 tests/gem_dummy_reloc_loop.c    |  107 +++++++++++++++++++++++-
 tests/gem_multi_bsd_sync_loop.c |  175 +++++++++++++++++++++++++++++++++++++++
 4 files changed, 283 insertions(+), 1 deletion(-)
 create mode 100644 tests/gem_multi_bsd_sync_loop.c

-- 
1.7.10.1

[PATCH I-g-t V4 1/2] tests: Add one ring sync case based on multi drm_fd to test ring semaphore sync under multi BSD rings

[Zhao Yakui]

The Broadwell GT3 machine has two independent BSD rings in kernel driver while
it is transparent to the user-space driver. In such case it needs to check
the ring sync between the two BSD rings. At the same time it also needs to
check the sync among the second BSD ring and the other rings.

V2->V3: Follow Imre's comment to remove the unnecessary initialization and
use igt_assert_f instead of igt_assert.

V3->V4: Add gem_multi_bsd_sync_loop.c into the tests/.gitignore

Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
---
 tests/.gitignore                |    1 +
 tests/Makefile.sources          |    1 +
 tests/gem_multi_bsd_sync_loop.c |  175 +++++++++++++++++++++++++++++++++++++++
 3 files changed, 177 insertions(+)
 create mode 100644 tests/gem_multi_bsd_sync_loop.c

diff --git a/tests/.gitignore b/tests/.gitignore
index 146bab0..42690dd 100644
--- a/tests/.gitignore
+++ b/tests/.gitignore
@@ -54,6 +54,7 @@ gem_media_fill
 gem_mmap
 gem_mmap_gtt
 gem_mmap_offset_exhaustion
+gem_multi_bsd_sync_loop
 gem_non_secure_batch
 gem_partial_pwrite_pread
 gem_persistent_relocs
diff --git a/tests/Makefile.sources b/tests/Makefile.sources
index c957ace..7cd9ca8 100644
--- a/tests/Makefile.sources
+++ b/tests/Makefile.sources
@@ -105,6 +105,7 @@ TESTS_progs = \
 	gem_render_tiled_blits \
 	gem_ring_sync_copy \
 	gem_ring_sync_loop \
+	gem_multi_bsd_sync_loop \
 	gem_seqno_wrap \
 	gem_set_tiling_vs_gtt \
 	gem_set_tiling_vs_pwrite \
diff --git a/tests/gem_multi_bsd_sync_loop.c b/tests/gem_multi_bsd_sync_loop.c
new file mode 100644
index 0000000..b01764a
--- /dev/null
+++ b/tests/gem_multi_bsd_sync_loop.c
@@ -0,0 +1,175 @@
+/*
+ * Copyright © 2014 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:
+ *    Daniel Vetter <daniel.vetter@ffwll.ch> (based on gem_ring_sync_loop_*.c)
+ *    Zhao Yakui <yakui.zhao@intel.com>
+ *
+ */
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <fcntl.h>
+#include <inttypes.h>
+#include <errno.h>
+#include <sys/stat.h>
+#include <sys/time.h>
+#include "drm.h"
+#include "ioctl_wrappers.h"
+#include "drmtest.h"
+#include "intel_bufmgr.h"
+#include "intel_batchbuffer.h"
+#include "intel_io.h"
+#include "i830_reg.h"
+#include "intel_chipset.h"
+
+static drm_intel_bufmgr *bufmgr;
+struct intel_batchbuffer *batch;
+static drm_intel_bo *target_buffer;
+
+#define NUM_FD	50
+
+static int mfd[NUM_FD];
+static drm_intel_bufmgr *mbufmgr[NUM_FD];
+static struct intel_batchbuffer *mbatch[NUM_FD];
+static drm_intel_bo *mbuffer[NUM_FD];
+
+
+/*
+ * Testcase: Basic check of ring<->ring sync using a dummy reloc
+ *
+ * Extremely efficient at catching missed irqs with semaphores=0 ...
+ */
+
+#define MI_COND_BATCH_BUFFER_END	(0x36<<23 | 1)
+#define MI_DO_COMPARE			(1<<21)
+
+static void
+store_dword_loop(int fd)
+{
+	int i;
+	int num_rings = gem_get_num_rings(fd);
+
+	srandom(0xdeadbeef);
+
+	for (i = 0; i < SLOW_QUICK(0x100000, 10); i++) {
+		int ring, mindex;
+		ring = random() % num_rings + 1;
+		mindex = random() % NUM_FD;
+		batch = mbatch[mindex];
+		if (ring == I915_EXEC_RENDER) {
+			BEGIN_BATCH(4);
+			OUT_BATCH(MI_COND_BATCH_BUFFER_END | MI_DO_COMPARE);
+			OUT_BATCH(0xffffffff); /* compare dword */
+			OUT_RELOC(mbuffer[mindex], I915_GEM_DOMAIN_RENDER,
+					I915_GEM_DOMAIN_RENDER, 0);
+			OUT_BATCH(MI_NOOP);
+			ADVANCE_BATCH();
+		} else {
+			BEGIN_BATCH(4);
+			OUT_BATCH(MI_FLUSH_DW | 1);
+			OUT_BATCH(0); /* reserved */
+			OUT_RELOC(mbuffer[mindex], I915_GEM_DOMAIN_RENDER,
+					I915_GEM_DOMAIN_RENDER, 0);
+			OUT_BATCH(MI_NOOP | (1<<22) | (0xf));
+			ADVANCE_BATCH();
+		}
+		intel_batchbuffer_flush_on_ring(batch, ring);
+	}
+
+	drm_intel_bo_map(target_buffer, 0);
+	// map to force waiting on rendering
+	drm_intel_bo_unmap(target_buffer);
+}
+
+igt_simple_main
+{
+	int fd;
+	int devid;
+	int i;
+
+	fd = drm_open_any();
+	devid = intel_get_drm_devid(fd);
+	gem_require_ring(fd, I915_EXEC_BLT);
+
+
+	bufmgr = drm_intel_bufmgr_gem_init(fd, 4096);
+	igt_assert_f(bufmgr, "fail to initialize the buf manager\n");
+	drm_intel_bufmgr_gem_enable_reuse(bufmgr);
+
+
+	target_buffer = drm_intel_bo_alloc(bufmgr, "target bo", 4096, 4096);
+	igt_assert_f(target_buffer, "fail to create the gem bo\n");
+
+	/* Create multiple drm_fd and map one gem_object among multi drm_fd */
+	{
+		unsigned int target_flink;
+		char buffer_name[32];
+		if (dri_bo_flink(target_buffer, &target_flink)) {
+			printf("fail to get flink for target buffer\n");
+			igt_assert_f(0, "fail to create global gem handle"
+				     " for allocated buffer\n");
+			goto fail_flink;
+		}
+		for (i = 0; i < NUM_FD; i++) {
+			sprintf(buffer_name, "Target buffer %d\n", i);
+			mfd[i] = drm_open_any();
+			mbufmgr[i] = drm_intel_bufmgr_gem_init(mfd[i], 4096);
+			igt_assert_f(mbufmgr[i],
+				     "fail to initialize buf manager for drm_fd %d\n",
+				     mfd[i]);
+			drm_intel_bufmgr_gem_enable_reuse(mbufmgr[i]);
+			mbatch[i] = intel_batchbuffer_alloc(mbufmgr[i], devid);
+			igt_assert_f(mbatch[i],
+				     "fail to create batchbuffer for drm_fd %d\n",
+				     mfd[i]);
+			mbuffer[i] = intel_bo_gem_create_from_name(mbufmgr[i], buffer_name, target_flink);
+			igt_assert_f(mbuffer[i],
+				     "fail to create buffer bo from global "
+				     "gem handle %d for drm_fd %d\n",
+                                     target_flink, mfd[i]);
+		}
+	}
+
+	store_dword_loop(fd);
+
+	{
+		for (i = 0; i < NUM_FD; i++) {
+			dri_bo_unreference(mbuffer[i]);
+			intel_batchbuffer_free(mbatch[i]);
+			drm_intel_bufmgr_destroy(mbufmgr[i]);
+			close(mfd[i]);
+		}
+	}
+	drm_intel_bo_unreference(target_buffer);
+	drm_intel_bufmgr_destroy(bufmgr);
+
+	close(fd);
+	return;
+
+fail_flink:
+	drm_intel_bo_unreference(target_buffer);
+	drm_intel_bufmgr_destroy(bufmgr);
+
+	close(fd);
+}
-- 
1.7.10.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH I-g-t V4 2/2] tests/gem_dummy_reloc_loop: Add one subtest based on multi drm_fd to test CPU<->GPU sync under multi BSD rings

[Zhao Yakui]

The Broadwell GT3 machine has two independent BSD rings in kernel driver while
it is transparent to the user-space driver. In such case it needs to check
the CPU<->GPU sync for the second BSD ring.

V1->V2: Follow Daniel's comment to add one subtext instead of one individual
test case, which is used to test the CPU<->GPU sync under multi BSD rings.

V2->V3: Follow Imre's comment to remove the unnecessary initialization and
use igt_assert_f instead of igt_assert

Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
---
 tests/gem_dummy_reloc_loop.c |  107 +++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 106 insertions(+), 1 deletion(-)

diff --git a/tests/gem_dummy_reloc_loop.c b/tests/gem_dummy_reloc_loop.c
index a61b59b..4e4dd49 100644
--- a/tests/gem_dummy_reloc_loop.c
+++ b/tests/gem_dummy_reloc_loop.c
@@ -48,6 +48,13 @@ static drm_intel_bufmgr *bufmgr;
 struct intel_batchbuffer *batch;
 static drm_intel_bo *target_buffer;
 
+#define NUM_FD	50
+
+static int mfd[NUM_FD];
+static drm_intel_bufmgr *mbufmgr[NUM_FD];
+static struct intel_batchbuffer *mbatch[NUM_FD];
+static drm_intel_bo *mbuffer[NUM_FD];
+
 /*
  * Testcase: Basic check of ring<->cpu sync using a dummy reloc
  *
@@ -124,6 +131,50 @@ dummy_reloc_loop_random_ring(int num_rings)
 	}
 }
 
+static void
+dummy_reloc_loop_random_ring_multi_fd(int num_rings)
+{
+	int i;
+	struct intel_batchbuffer *saved_batch;
+
+	saved_batch = batch;
+
+	srandom(0xdeadbeef);
+
+	for (i = 0; i < 0x100000; i++) {
+		int mindex;
+		int ring = random() % num_rings + 1;
+
+		mindex = random() % NUM_FD;
+		batch = mbatch[mindex];
+
+		if (ring == I915_EXEC_RENDER) {
+			BEGIN_BATCH(4);
+			OUT_BATCH(MI_COND_BATCH_BUFFER_END | MI_DO_COMPARE);
+			OUT_BATCH(0xffffffff); /* compare dword */
+			OUT_RELOC(mbuffer[mindex], I915_GEM_DOMAIN_RENDER,
+					I915_GEM_DOMAIN_RENDER, 0);
+			OUT_BATCH(MI_NOOP);
+			ADVANCE_BATCH();
+		} else {
+			BEGIN_BATCH(4);
+			OUT_BATCH(MI_FLUSH_DW | 1);
+			OUT_BATCH(0); /* reserved */
+			OUT_RELOC(mbuffer[mindex], I915_GEM_DOMAIN_RENDER,
+					I915_GEM_DOMAIN_RENDER, 0);
+			OUT_BATCH(MI_NOOP | (1<<22) | (0xf));
+			ADVANCE_BATCH();
+		}
+		intel_batchbuffer_flush_on_ring(batch, ring);
+
+		drm_intel_bo_map(target_buffer, 0);
+		// map to force waiting on rendering
+		drm_intel_bo_unmap(target_buffer);
+	}
+
+	batch = saved_batch;
+}
+
 int fd;
 int devid;
 int num_rings;
@@ -133,6 +184,7 @@ igt_main
 	igt_skip_on_simulation();
 
 	igt_fixture {
+		int i;
 		fd = drm_open_any();
 		devid = intel_get_drm_devid(fd);
 		num_rings = gem_get_num_rings(fd);
@@ -148,6 +200,40 @@ igt_main
 
 		target_buffer = drm_intel_bo_alloc(bufmgr, "target bo", 4096, 4096);
 		igt_assert(target_buffer);
+
+		/* Create multi drm_fd and map one gem object to multi gem_contexts */
+		{
+			unsigned int target_flink;
+			char buffer_name[32];
+			if (dri_bo_flink(target_buffer, &target_flink)) {
+				printf("fail to get flink for target buffer\n");
+				igt_assert_f(0, "fail to create global "
+					     "gem_handle for target buffer\n");
+			}
+			for (i = 0; i < NUM_FD; i++) {
+				sprintf(buffer_name, "Target buffer %d\n", i);
+				mfd[i] = drm_open_any();
+				mbufmgr[i] = drm_intel_bufmgr_gem_init(mfd[i], 4096);
+				igt_assert_f(mbufmgr[i],
+					     "fail to initialize buf manager "
+					     "for drm_fd %d\n",
+					     mfd[i]);
+				drm_intel_bufmgr_gem_enable_reuse(mbufmgr[i]);
+				mbatch[i] = intel_batchbuffer_alloc(mbufmgr[i], devid);
+				igt_assert_f(mbatch[i],
+					     "fail to create batchbuffer "
+					     "for drm_fd %d\n",
+					     mfd[i]);
+				mbuffer[i] = intel_bo_gem_create_from_name(
+								mbufmgr[i],
+								buffer_name,
+								target_flink);
+				igt_assert_f(mbuffer[i],
+					     "fail to create gem bo from global "
+					     "gem_handle %d for drm_fd %d\n",
+					     target_flink, mfd[i]);
+			}
+		}
 	}
 
 	igt_subtest("render") {
@@ -190,8 +276,27 @@ igt_main
 			printf("dummy loop run on random rings completed\n");
 		}
 	}
-
+	igt_subtest("mixed_multi_fd") {
+		if (num_rings > 1) {
+			sleep(2);
+			printf("running dummy loop on random rings based on "
+					"multi drm_fd\n");
+			dummy_reloc_loop_random_ring_multi_fd(num_rings);
+			printf("dummy loop run on random rings based on "
+					"multi drm_fd completed\n");
+		}
+	}
 	igt_fixture {
+		int i;
+		/* Free the buffer/batchbuffer/buffer mgr for multi-fd */
+		{
+			for (i = 0; i < NUM_FD; i++) {
+				dri_bo_unreference(mbuffer[i]);
+				intel_batchbuffer_free(mbatch[i]);
+				drm_intel_bufmgr_destroy(mbufmgr[i]);
+				close(mfd[i]);
+			}
+		}
 		drm_intel_bo_unreference(target_buffer);
 		intel_batchbuffer_free(batch);
 		drm_intel_bufmgr_destroy(bufmgr);
-- 
1.7.10.1

Re: [PATCH I-g-t V4 0/2] Tests: Add test cases based on multi drm_fd to test sync

[Daniel Vetter]

On Thu, Apr 24, 2014 at 10:15:43AM +0800, Zhao Yakui wrote:
> This follows Daniel's advice to add the two test cases based on multi drm_fd to 
> test the ring sync and CPU<->GPU sync.
> The Broadwell GT3 machine has two independent BSD rings that can be used
> to process the video commands. This is implemented in kernel driver and transparent
> to the user-space. But we still need to check the ring sync and CPU<->GPU sync for
> the second BSD ring. Two tests are created based on the multi drm_fds to
> test the sync. Multi drm_fd can assure that the second BSD ring has the opportunity
> to dispatch the GPU command. 
> 
> V1->V2: Follow Daniel's comment to add one subtext instead of one individual
> test case, which is used to test the CPU<->GPU sync under multi BSD rings/
> 
> V2->V3: Follow Imre's comment to remove the unnecessary initialization and
> use igt_assert_f instead of igt_assert
> 
> V3->V4: Add gem_multi_bsd_sync_loop.c into the tests/.gitignore
> 
> Zhao Yakui (2):
>   tests: Add one ring sync case based on multi drm_fd to test ring
>     semaphore sync under multi BSD rings
>   tests/gem_dummy_reloc_loop: Add one subtest based on multi drm_fd to
>     test CPU<->GPU sync under multi BSD rings

Merged, thanks for the patches.
-Daniel

> 
>  tests/.gitignore                |    1 +
>  tests/Makefile.sources          |    1 +
>  tests/gem_dummy_reloc_loop.c    |  107 +++++++++++++++++++++++-
>  tests/gem_multi_bsd_sync_loop.c |  175 +++++++++++++++++++++++++++++++++++++++
>  4 files changed, 283 insertions(+), 1 deletion(-)
>  create mode 100644 tests/gem_multi_bsd_sync_loop.c
> 
> -- 
> 1.7.10.1
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx

-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch

Re: [PATCH I-g-t V4 0/2] Tests: Add test cases based on multi drm_fd to test sync

[Zhao Yakui]

On Wed, 2014-04-23 at 20:02 -0600, Zhao, Yakui wrote:

It seems that the patch 01 is filter out.
So I will try to resend it again.

Thanks.
    Yakui

> This follows Daniel's advice to add the two test cases based on multi drm_fd to 
> test the ring sync and CPU<->GPU sync.
> The Broadwell GT3 machine has two independent BSD rings that can be used
> to process the video commands. This is implemented in kernel driver and transparent
> to the user-space. But we still need to check the ring sync and CPU<->GPU sync for
> the second BSD ring. Two tests are created based on the multi drm_fds to
> test the sync. Multi drm_fd can assure that the second BSD ring has the opportunity
> to dispatch the GPU command. 
> 
> V1->V2: Follow Daniel's comment to add one subtext instead of one individual
> test case, which is used to test the CPU<->GPU sync under multi BSD rings/
> 
> V2->V3: Follow Imre's comment to remove the unnecessary initialization and
> use igt_assert_f instead of igt_assert
> 
> 
> Zhao Yakui (2):
>   tests: Add one ring sync case based on multi drm_fd to test ring
>     semaphore sync under multi BSD rings
>   tests/gem_dummy_reloc_loop: Add one subtest based on multi drm_fd to
>     test CPU<->GPU sync under multi BSD rings
> 
>  tests/.gitignore                |    1 +
>  tests/Makefile.sources          |    1 +
>  tests/gem_dummy_reloc_loop.c    |  107 +++++++++++++++++++++++-
>  tests/gem_multi_bsd_sync_loop.c |  175 +++++++++++++++++++++++++++++++++++++++
>  4 files changed, 283 insertions(+), 1 deletion(-)
>  create mode 100644 tests/gem_multi_bsd_sync_loop.c
> 

[PATCH I-g-t V4 0/2] Tests: Add test cases based on multi drm_fd to test sync

[Zhao Yakui]

This follows Daniel's advice to add the two test cases based on multi drm_fd to 
test the ring sync and CPU<->GPU sync.
The Broadwell GT3 machine has two independent BSD rings that can be used
to process the video commands. This is implemented in kernel driver and transparent
to the user-space. But we still need to check the ring sync and CPU<->GPU sync for
the second BSD ring. Two tests are created based on the multi drm_fds to
test the sync. Multi drm_fd can assure that the second BSD ring has the opportunity
to dispatch the GPU command. 

V1->V2: Follow Daniel's comment to add one subtext instead of one individual
test case, which is used to test the CPU<->GPU sync under multi BSD rings/

V2->V3: Follow Imre's comment to remove the unnecessary initialization and
use igt_assert_f instead of igt_assert


Zhao Yakui (2):
  tests: Add one ring sync case based on multi drm_fd to test ring
    semaphore sync under multi BSD rings
  tests/gem_dummy_reloc_loop: Add one subtest based on multi drm_fd to
    test CPU<->GPU sync under multi BSD rings

 tests/.gitignore                |    1 +
 tests/Makefile.sources          |    1 +
 tests/gem_dummy_reloc_loop.c    |  107 +++++++++++++++++++++++-
 tests/gem_multi_bsd_sync_loop.c |  175 +++++++++++++++++++++++++++++++++++++++
 4 files changed, 283 insertions(+), 1 deletion(-)
 create mode 100644 tests/gem_multi_bsd_sync_loop.c

-- 
1.7.10.1