[igt-dev] [PATCH i-g-t v8 3/6] tests/i915_pm_lpsp: lpsp platform agnostic support

Petri Latvala petri.latvala at intel.com
Tue May 19 09:32:09 UTC 2020


On Tue, May 12, 2020 at 04:04:12PM +0530, Anshuman Gupta wrote:
> Current implementation of lpsp igt test, assumed that every non-edp
> panel isn't a lpsp panel but it is not true on TGL anymore,
> any HDMI/DP/DSI panel connected on pipe A and connected to PORT_{A,B,C}
> can drive LPSP.
> Even on older Gen9 platform a DP panel can drive lpsp on Port A.
> This requires complete design change in current lpsp igt for a platform
> agnostic support.
> 
> The new igt approach is relies on connector specific
> i915_lpsp_capability and i915_lpsp_status debugfs attributes,
> these debugfs exposes whether an output is capable of driving lpsp
> and lpsp is enabled.
> 
> Nuking edp-native and non-edp test, introducing a new dynamic
> igt subtest kms-lpsp, which validates lpsp on each connected output
> and skip the subtest if output is not lpsp capable.
> 
> Skip screens-disabled test for platform which support DC states.
> screens-disabled test is only valid for the platforms like HSW/BDW
> where PG1 is Always-ON power domain, so these platform test lpsp
> with all screen disabled i.e validate PG2 when all screen are
> disabled.
> DC state i.e DMC f/w supported platforms don't have any ROI to validate
> screens-disabled subtest as existing dc*-dpms i915_pm_dc
> subtest already validate it implicitly.
> 
> v2:
> - CI failures fixup.
> v3:
> - removed unloading of snd_modules. [Martin]
> v4:
> - Don't test non-lpsp(if lpsp disabled), no ROI to test that.
> - nuke panel-fitter test.
> v5:
> - Added dynamic subtest and igt changes according to kernel
>   debugfs i915_lpsp_status changes.
> v6:
> - Avoided pipe big joiner by using 4k or lesser mode,
>   so igt will not fail when pipe big joiner patches will
>   land to kernel.
> v7:
> - Combined [v6 4/6] patch to this patch, skip the screens-disabled
>   subtest for DC state supported platforms. [Animesh]
> v8:
> - 4k mode correction 3840*2160 and for loop for available modes
>   should iterate from c->modes[0]. [Ankit]
> 
> Reviewed-by: Animesh Manna <animesh.manna at intel.com>
> Signed-off-by: Anshuman Gupta <anshuman.gupta at intel.com>
> ---
>  tests/i915/i915_pm_lpsp.c | 313 ++++++++++++++++++--------------------
>  1 file changed, 147 insertions(+), 166 deletions(-)
> 
> diff --git a/tests/i915/i915_pm_lpsp.c b/tests/i915/i915_pm_lpsp.c
> index 08f82e7c..0e9452fd 100644
> --- a/tests/i915/i915_pm_lpsp.c
> +++ b/tests/i915/i915_pm_lpsp.c
> @@ -25,210 +25,191 @@
>   */
>  
>  #include "igt.h"
> +#include "igt_kmod.h"
> +#include "igt_pm.h"
> +#include "igt_sysfs.h"
>  #include <sys/types.h>
>  #include <sys/stat.h>
>  #include <fcntl.h>
>  #include <unistd.h>
>  
> +#define MAX_SINK_LPSP_INFO_BUF_LEN	4096
>  
> -static bool supports_lpsp(uint32_t devid)
> -{
> -	return IS_HASWELL(devid) || IS_BROADWELL(devid);
> -}
> +#define PWR_DOMAIN_INFO "i915_power_domain_info"
>  
> -static bool lpsp_is_enabled(int drm_fd)
> +typedef struct {
> +	int drm_fd;
> +	int debugfs_fd;
> +	uint32_t devid;
> +	char *pwr_dmn_info;
> +	igt_display_t display;
> +	struct igt_fb fb;
> +	drmModeModeInfo *mode;
> +	igt_output_t *output;
> +} data_t;
> +
> +static bool lpsp_is_enabled(data_t *data)
>  {
> -	uint32_t val;
> +	char buf[MAX_SINK_LPSP_INFO_BUF_LEN];
> +	int len;
>  
> -	val = INREG(HSW_PWR_WELL_CTL2);
> -	return !(val & HSW_PWR_WELL_STATE_ENABLED);
> -}
> +	len = igt_debugfs_simple_read(data->debugfs_fd, "i915_lpsp_status",
> +				      buf, sizeof(buf));
> +	if (len < 0)
> +		igt_assert_eq(len, -ENODEV);
>  
> -/* The LPSP mode is all about an enabled pipe, but we expect to also be in the
> - * low power mode when no pipes are enabled, so do this check anyway. */
> -static void screens_disabled_subtest(int drm_fd, drmModeResPtr drm_res)
> -{
> -	kmstest_unset_all_crtcs(drm_fd, drm_res);
> -	igt_assert(lpsp_is_enabled(drm_fd));
> +	igt_skip_on(strstr(buf, "LPSP: not supported"));
> +
> +	return strstr(buf, "LPSP: enabled");
>  }
>  
> -static uint32_t create_fb(int drm_fd, int width, int height)
> +static bool dmc_supported(int debugfs)
>  {
> -	struct igt_fb fb;
> +	char buf[15];
> +	int len;
>  
> -	return igt_create_pattern_fb(drm_fd, width, height, DRM_FORMAT_XRGB8888,
> -				     LOCAL_DRM_FORMAT_MOD_NONE, &fb);
> +	len = igt_sysfs_read(debugfs, "i915_dmc_info", buf, sizeof(buf) - 1);
> +
> +	if (len < 0)
> +		return false;
> +	else
> +		return true;
>  }
>  
> -static void edp_subtest(int drm_fd, drmModeResPtr drm_res,
> -			drmModeConnectorPtr *drm_connectors, uint32_t devid,
> -			bool use_panel_fitter)
> +/*
> + * The LPSP mode is all about an enabled pipe, but we expect to also be in the
> + * low power mode when no pipes are enabled, so do this check anyway.
> + */
> +static void screens_disabled_subtest(data_t *data)
>  {
> -	int i, rc;
> -	uint32_t crtc_id = 0, buffer_id = 0;
> -	drmModeConnectorPtr connector = NULL;
> -	drmModeModeInfoPtr mode = NULL;
> -	drmModeModeInfo std_1024_mode = {
> -		.clock = 65000,
> -		.hdisplay = 1024,
> -		.hsync_start = 1048,
> -		.hsync_end = 1184,
> -		.htotal = 1344,
> -		.hskew = 0,
> -		.vdisplay = 768,
> -		.vsync_start = 771,
> -		.vsync_end = 777,
> -		.vtotal = 806,
> -		.vscan = 0,
> -		.vrefresh = 60,
> -		.flags = 0xA,
> -		.type = 0x40,
> -		.name = "Custom 1024x768",
> -	};
> -
> -	kmstest_unset_all_crtcs(drm_fd, drm_res);
> -
> -	for (i = 0; i < drm_res->count_connectors; i++) {
> -		drmModeConnectorPtr c = drm_connectors[i];
> -
> -		if (c->connector_type != DRM_MODE_CONNECTOR_eDP)
> -			continue;
> -		if (c->connection != DRM_MODE_CONNECTED)
> -			continue;
> -
> -		if (!use_panel_fitter && c->count_modes) {
> -			connector = c;
> -			mode = &c->modes[0];
> -			break;
> -		}
> -		if (use_panel_fitter) {
> -			connector = c;
> -
> -			/* This is one of the modes Xorg creates for panels, so
> -			 * it should work just fine. Notice that Gens that
> -			 * support LPSP are too new for panels with native
> -			 * 1024x768 resolution, so this should force the panel
> -			 * fitter. */
> -			igt_assert(c->count_modes &&
> -				   c->modes[0].hdisplay > 1024);
> -			igt_assert(c->count_modes &&
> -				   c->modes[0].vdisplay > 768);
> -			mode = &std_1024_mode;
> -			break;
> -		}
> -	}
> -	igt_require(connector);
> -
> -	crtc_id = kmstest_find_crtc_for_connector(drm_fd, drm_res, connector,
> -						  0);
> -	buffer_id = create_fb(drm_fd, mode->hdisplay, mode->vdisplay);
> -
> -	igt_assert(buffer_id);
> -	igt_assert(connector);
> -	igt_assert(mode);
> -
> -	rc = drmModeSetCrtc(drm_fd, crtc_id, buffer_id, 0, 0,
> -			    &connector->connector_id, 1, mode);
> -	igt_assert_eq(rc, 0);
> -
> -	if (use_panel_fitter) {
> -		if (IS_HASWELL(devid))
> -			igt_assert(!lpsp_is_enabled(drm_fd));
> -		else
> -			igt_assert(lpsp_is_enabled(drm_fd));
> -	} else {
> -		igt_assert(lpsp_is_enabled(drm_fd));
> +	igt_output_t *output;
> +	int valid_output = 0;
> +	enum pipe pipe;
> +
> +	for_each_pipe_with_single_output(&data->display, pipe, output) {
> +		data->output = output;
> +		igt_output_set_pipe(data->output, PIPE_NONE);
> +		igt_display_commit(&data->display);
> +		valid_output++;
>  	}

This code is already merged, but nonetheless:

This for loop doesn't do what you probably think it does. This loops
through each pipe, possibly giving you the same output for every pipe.

An immediate effect is that this introduces a new compile warning:

../tests/i915/i915_pm_lpsp.c: In function ‘screens_disabled_subtest’:
../tests/i915/i915_pm_lpsp.c:87:12: warning: variable ‘pipe’ set but not used [-Wunused-but-set-variable]
  enum pipe pipe;
              ^~~~



-- 
Petri Latvala


More information about the igt-dev mailing list