From: Dibin Moolakadan Subrahmanian <dibin.moolakadan.subrahmanian@intel.com>
To: Jeevan B <jeevan.b@intel.com>, <igt-dev@lists.freedesktop.org>
Cc: <ramanaidu.naladala@intel.com>
Subject: Re: [PATCH i-g-t] tests/intel/kms_pm_dc: Detect frame drops via page-flip events
Date: Mon, 20 Jul 2026 10:49:27 +0530 [thread overview]
Message-ID: <db573a55-2f2b-46d3-a020-22c81d6bed10@intel.com> (raw)
In-Reply-To: <20260709050905.4077066-1-jeevan.b@intel.com>
On 7/9/2026 10:39 AM, Jeevan B wrote:
> The DC3CO frame-drop test currently waits for a vblank after committing
> a new framebuffer, which does not guarantee that the frame was actually
> presented.
>
> Switch to page flips and wait for the corresponding completion event,
> which confirms that the frame reached scanout. The test now alternates
> between two framebuffers using page flips, preserving the playback-like
> cadence needed to exercise DC3CO entry/exit transitions.
>
> Keep the existing vblank-gap validation, but derive the sequence number
> from the completed page flip instead of an independently waited vblank.
>
> Signed-off-by: Jeevan B <jeevan.b@intel.com>
> ---
> tests/intel/kms_pm_dc.c | 69 +++++++++++++++++++++++++++++++----------
> 1 file changed, 52 insertions(+), 17 deletions(-)
>
> diff --git a/tests/intel/kms_pm_dc.c b/tests/intel/kms_pm_dc.c
> index ef53c5f96..da077a2dc 100644
> --- a/tests/intel/kms_pm_dc.c
> +++ b/tests/intel/kms_pm_dc.c
> @@ -362,30 +362,60 @@ static void test_dc3co_vpb_simulation(data_t *data)
> cleanup_dc3co_fbs(data);
> }
>
> -static uint32_t wait_for_next_vblank_seq(data_t *data)
> +struct dc3co_flip {
> + bool done;
> + unsigned int seq;
> +};
> +
> +static void dc3co_flip_handler(int fd, unsigned int seq, unsigned int tv_sec,
> + unsigned int tv_usec, void *user_data)
> {
> - drmVBlank wait = {};
> - igt_crtc_t *crtc = data->output->pending_crtc;
> + struct dc3co_flip *flip = user_data;
>
> - igt_assert_f(crtc, "No CRTC bound to output for vblank wait\n");
> + flip->seq = seq;
> + flip->done = true;
> +}
>
> - wait.request.type = kmstest_get_vbl_flag(crtc->crtc_index) |
> - DRM_VBLANK_RELATIVE |
> - DRM_VBLANK_NEXTONMISS;
> - wait.request.sequence = 1;
> - igt_assert_eq(drmWaitVBlank(data->drm_fd, &wait), 0);
> +static unsigned int dc3co_flip_and_wait(data_t *data, uint32_t crtc_id,
> + uint32_t fb_id, int frame_time_us)
> +{
> + drmEventContext evctx = {
> + .version = 2,
> + .page_flip_handler = dc3co_flip_handler,
> + };
> + struct dc3co_flip flip = {};
> + struct pollfd pfd = {
> + .fd = data->drm_fd,
> + .events = POLLIN,
> + };
> + int timeout_ms = (DC3CO_MAX_VBLANK_GAP + 2) * frame_time_us / 1000;
> +
> + if (timeout_ms < 50)
> + timeout_ms = 50;
>
> - return wait.reply.sequence;
> + igt_assert_eq(drmModePageFlip(data->drm_fd, crtc_id, fb_id,
> + DRM_MODE_PAGE_FLIP_EVENT, &flip), 0);
> +
> + while (!flip.done) {
> + igt_assert_f(poll(&pfd, 1, timeout_ms) > 0,
> + "Frame dropped: no page flip completion within %d ms "
> + "during DC3CO entry/exit\n", timeout_ms);
> + igt_assert_eq(drmHandleEvent(data->drm_fd, &evctx), 0);
> + }
> +
> + return flip.seq;
> }
>
> static void detect_dc3co_framedrop(data_t *data)
> {
> igt_plane_t *primary;
> + igt_crtc_t *crtc;
> uint32_t dc3co_prev_cnt;
> uint32_t dc3co_cur_cnt;
> uint32_t prev_vblank_seq = 0;
> uint32_t vblank_seq;
> uint32_t vblank_gap;
> + int frame_time_us;
> int delay;
> int committed = 0;
> bool dc3co_after_target = false;
> @@ -393,25 +423,30 @@ static void detect_dc3co_framedrop(data_t *data)
>
> igt_require_f(data->mode->vrefresh != 0, "Invalid vrefresh rate of 0\n");
>
> + crtc = data->output->pending_crtc;
> + igt_assert_f(crtc, "No CRTC bound to output\n");
> +
> primary = igt_output_get_plane_type(data->output, DRM_PLANE_TYPE_PRIMARY);
> - igt_plane_set_fb(primary, NULL);
> +
> + igt_plane_set_fb(primary, &data->fb_rgb);
> igt_display_commit(&data->display);
>
> dc3co_prev_cnt = igt_read_dc_counter(data->debugfs_fd, IGT_INTEL_CHECK_DC3CO);
>
> - delay = (int)(DC3CO_FRAME_DELAY_FACTOR * (1000000 / data->mode->vrefresh));
> + frame_time_us = 1000000 / data->mode->vrefresh;
> + delay = (int)(DC3CO_FRAME_DELAY_FACTOR * frame_time_us);
>
> while (committed < DC3CO_VERIFY_COMMITS) {
> front = !front;
> - igt_plane_set_fb(primary, front ? &data->fb_rgr : &data->fb_rgb);
> - igt_display_commit(&data->display);
> -
> - vblank_seq = wait_for_next_vblank_seq(data);
> + vblank_seq = dc3co_flip_and_wait(data, crtc->crtc_id,
> + front ? data->fb_rgr.fb_id :
> + data->fb_rgb.fb_id,
> + frame_time_us);
> if (prev_vblank_seq) {
> vblank_gap = vblank_seq - prev_vblank_seq;
> igt_assert_f(igt_vblank_after(vblank_seq, prev_vblank_seq) &&
> vblank_gap <= DC3CO_MAX_VBLANK_GAP,
> - "Unexpected vblank gap %u after commit %d (prev=%u, cur=%u)\n",
> + "Frame drop: unexpected vblank gap %u after commit %d (prev=%u, cur=%u)\n",
> vblank_gap, committed + 1, prev_vblank_seq, vblank_seq);
> }
> prev_vblank_seq = vblank_seq;
Page flips fit video-playback simulation better than the vblank wait.
LGTM.
Reviewed-by: Dibin Moolakadan Subrahmanian <dibin.moolakadan.subrahmanian@intel.com>
prev parent reply other threads:[~2026-07-20 5:20 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-07-09 5:09 [PATCH i-g-t] tests/intel/kms_pm_dc: Detect frame drops via page-flip events Jeevan B
2026-07-09 7:22 ` ✓ Xe.CI.BAT: success for " Patchwork
2026-07-09 7:55 ` ✓ i915.CI.BAT: " Patchwork
2026-07-09 10:44 ` ✓ Xe.CI.FULL: " Patchwork
2026-07-10 2:40 ` ✗ i915.CI.Full: failure " Patchwork
2026-07-20 5:19 ` Dibin Moolakadan Subrahmanian [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=db573a55-2f2b-46d3-a020-22c81d6bed10@intel.com \
--to=dibin.moolakadan.subrahmanian@intel.com \
--cc=igt-dev@lists.freedesktop.org \
--cc=jeevan.b@intel.com \
--cc=ramanaidu.naladala@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox