[Intel-gfx] [PATCH i-g-t] tests/kms_atomic_abi: Test event ABI corner cases

Maarten Lankhorst maarten.lankhorst at linux.intel.com
Wed May 3 12:04:24 UTC 2017


Op 27-04-17 om 14:42 schreef Mika Kahola:
> Atomic has a few special cases around async commits and event generation
> that we need to test. This patch addresses these two tests
>
> - kernel rejects events on a disabled pipe
> - events on a pipe that is getting enabled/disabled
>
> For: VIZ-6954
>
> Signed-off-by: Mika Kahola <mika.kahola at intel.com>
> ---
>  tests/Makefile.sources |   1 +
>  tests/kms_atomic_abi.c | 201 +++++++++++++++++++++++++++++++++++++++++++++++++
>  2 files changed, 202 insertions(+)
>  create mode 100644 tests/kms_atomic_abi.c
>
> diff --git a/tests/Makefile.sources b/tests/Makefile.sources
> index 2b6e6ee..b4e9897 100644
> --- a/tests/Makefile.sources
> +++ b/tests/Makefile.sources
> @@ -121,6 +121,7 @@ TESTS_progs_M = \
>  	kms_plane \
>  	kms_plane_multiple \
>  	kms_plane_lowres \
> +	kms_atomic_abi \
>  	kms_properties \
>  	kms_psr_sink_crc \
>  	kms_render \
> diff --git a/tests/kms_atomic_abi.c b/tests/kms_atomic_abi.c
> new file mode 100644
> index 0000000..843220b
> --- /dev/null
> +++ b/tests/kms_atomic_abi.c
> @@ -0,0 +1,201 @@
> +/*
> + * Copyright © 2017 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.
> + *
> + */
> +
> +#include "igt.h"
> +#include <fcntl.h>
> +
> +IGT_TEST_DESCRIPTION("Test event ABI");
> +
> +static void
> +do_atomic_commit(igt_display_t *display, int expectation)
> +{
> +	char buf[256];
> +	struct drm_event *e = (void *)buf;
> +	int ret;
> +
> +	ret = igt_display_try_commit_atomic(display,
> +					    DRM_MODE_PAGE_FLIP_EVENT,
> +					    NULL);
> +	igt_assert_eq(ret, expectation);
> +
> +	if (expectation == 0) {

I know they're equal, but should be ret == 0 for clarity? :)

Might even be better for readability to remove this function entirely and add a bool to read_pageflip_event whether an event is expected and return error on failure. :)

> +		igt_set_timeout(1, "Stuck on page flip");
> +		ret = read(display->drm_fd, buf, sizeof(buf));
> +		igt_assert(ret >= 0);
> +
> +		igt_reset_timeout();
> +		igt_assert_eq(e->type, DRM_EVENT_FLIP_COMPLETE);
> +	}
> +}
> +
> +static int
> +read_pageflip_event(igt_display_t *display)
> +{
> +	char buf[256];
> +	struct drm_event *e = (void *)buf;
> +	int ret;
> +
> +	for (int i = 0; i < 100; i++) {
> +		ret = read(display->drm_fd, buf, sizeof(buf));
> +
> +		if (ret >= 0)
> +			break;
> +	}
I think just reading 100 times would cause the test to depend on how fast the system is. Might be better to use poll(50ms) + (optional) read() instead?
Would also mean the NONBLOCK fudging can be removed. :)
> +
> +	if (ret >= 0)
> +		igt_assert_eq(e->type, DRM_EVENT_FLIP_COMPLETE);
> +
> +	return ret;
> +}
> +
> +static void
> +test_init(igt_display_t *display, enum pipe pipe, igt_output_t *output)
> +{
> +	struct igt_fb fb;
> +	drmModeModeInfo *mode;
> +	igt_plane_t *primary;
> +	int flags = DRM_MODE_ATOMIC_ALLOW_MODESET | DRM_MODE_ATOMIC_NONBLOCK;
> +	int ret;
> +
> +	igt_output_set_pipe(output, pipe);
> +
> +	primary = igt_output_get_plane_type(output, DRM_PLANE_TYPE_PRIMARY);
> +
> +	mode = igt_output_get_mode(output);
> +
> +	igt_create_color_fb(display->drm_fd, mode->hdisplay, mode->vdisplay,
> +			    DRM_FORMAT_XRGB8888,
> +			    LOCAL_DRM_FORMAT_MOD_NONE,
> +			    0.0f, 0.0f, 1.0f,
> +			    &fb);
> +
> +	igt_plane_set_fb(primary, &fb);
> +
> +	ret = igt_display_try_commit_atomic(display, flags, NULL);
> +	igt_assert_eq(ret, 0);
> +}
modeset disable should generate a valid event too, as should modeset enable. Might test those here too? :)
> +
> +static void
> +test_disable_pipe(igt_display_t *display, enum pipe pipe, igt_output_t *output)
> +{
> +	test_init(display, pipe, output);
> +
> +	igt_output_set_pipe(output, pipe);
> +
> +	do_atomic_commit(display, 0);
> +
> +	/* disable pipe */
> +	igt_output_set_pipe(output, PIPE_NONE);
> +
> +	/* try to do atomic commit */
> +	do_atomic_commit(display, -EINVAL);
> +
> +	igt_output_set_pipe(output, PIPE_ANY);
> +}
> +
> +static void
> +test_dpms(igt_display_t *display, enum pipe pipe, igt_output_t *output)
> +{
> +	int ret;
> +	int flags = fcntl(display->drm_fd, F_GETFL, 0);
> +
> +	test_init(display, pipe, output);
> +
> +	igt_output_set_pipe(output, pipe);
> +
> +	do_atomic_commit(display, 0);
> +
> +	fcntl(display->drm_fd, F_SETFL, flags | O_NONBLOCK);
> +
> +	kmstest_set_connector_dpms(display->drm_fd, output->config.connector,
> +				   DRM_MODE_DPMS_OFF);
It might be worth adding the pipe active property that's reset to default on each call to igt_output_set_pipe,
so the default will be track pipe enabled, but with the option to override to always on or always off.
> +	/* try to do atomic commit */
> +	ret = igt_display_try_commit_atomic(display,
> +					    DRM_MODE_PAGE_FLIP_EVENT,
> +					    NULL);
> +	igt_assert_eq(ret, 0);
> +
> +	/* with pipe off, we shouldn't receive pageflip event */
> +	ret = read_pageflip_event(display);
> +	igt_assert_eq(ret, -1);
Either try_commit_atomic fails during DPMS off and we return -EINVAL, or we succeed and get an event here. I'm not sure this test works the way you expect it to.
Either way it's likely this test is broken here. :)
> +
> +	fcntl(display->drm_fd, F_SETFL, flags & ~O_NONBLOCK);
> +
> +	kmstest_set_connector_dpms(display->drm_fd, output->config.connector,
> +				   DRM_MODE_DPMS_ON);
> +
> +	igt_output_set_pipe(output, PIPE_ANY);
> +
> +	fcntl(display->drm_fd, F_SETFL, flags & ~O_NONBLOCK);
> +}
> +
> +igt_main
> +{
> +	igt_output_t *output;
> +	igt_display_t display;
> +	enum pipe pipe;
> +	int valid_tests;
> +
> +	igt_skip_on_simulation();
> +
> +	igt_fixture {
> +		display.drm_fd = drm_open_driver_master(DRIVER_ANY);
> +		kmstest_set_vt_graphics_mode();
> +		igt_display_init(&display, display.drm_fd);
> +		igt_require(display.is_atomic);
> +	}
> +
> +	igt_subtest_f("disabled-pipe") {
> +		valid_tests = 0;
> +		for_each_pipe_with_valid_output(&display, pipe, output) {
> +			test_disable_pipe(&display, pipe, output);
> +
> +			valid_tests++;
> +			break;
> +		}
> +
> +		igt_require_f(valid_tests, "no valid crtc/connector combinations found\n");
> +	}
> +
> +	igt_subtest_f("dpms") {
> +		valid_tests = 0;
> +		for_each_pipe_with_valid_output(&display, pipe, output) {
> +			test_dpms(&display, pipe, output);
> +
> +			valid_tests++;
> +			break;
> +		}
> +
> +		igt_require_f(valid_tests, "no valid crtc/connector combinations found\n");
> +	}
> +
> +	igt_fixture {
> +		igt_display_fini(&display);
> +		close(display.drm_fd);
> +	}
> +
> +	igt_exit();
> +}




More information about the Intel-gfx mailing list