Igt-dev Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: "Maíra Canal" <mcanal@igalia.com>
To: Iago Toral <itoral@igalia.com>, Melissa Wen <mwen@igalia.com>,
	Kamil Konieczny <kamil.konieczny@linux.intel.com>,
	Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com>,
	Bhanuprakash Modem <bhanuprakash.modem@gmail.com>,
	Ashutosh Dixit <ashutosh.dixit@intel.com>,
	Karthik B S <karthik.b.s@intel.com>
Cc: igt-dev@lists.freedesktop.org, kernel-dev@igalia.com,
	"Maíra Canal" <mcanal@igalia.com>
Subject: [PATCH i-g-t 0/5] tests/v3d_perfmon: Fix V3D 7.1 regression and extend perfmon testing
Date: Fri,  8 May 2026 09:42:49 -0300	[thread overview]
Message-ID: <20260508124446.1260672-2-mcanal@igalia.com> (raw)

This series extends the V3D perfmon coverage in two directions: it fixes
an existing test that regressed on V3D 7.1 (Raspberry Pi 5), and it adds
coverage for the new DRM_IOCTL_V3D_PERFMON_SET_GLOBAL ioctl plus companion
counter-isolation tests.

PATCH 1 fixes create-perfmon-invalid-counters, which currently fails on
V3D 7.1 because it hardcodes V3D_PERFCNT_NUM (only valid for V3D 4.2)
instead of querying DRM_V3D_PARAM_MAX_PERF_COUNTERS as the UAPI prescribes.

PATCHES 2-4 are the plumbing for the new tests:

  - sync the v3d UAPI header to expose DRM_IOCTL_V3D_PERFMON_SET_GLOBAL;
  - extend igt_v3d_perfmon_get_values() so callers can read counter values
		back into a caller-provided buffer;
  - add a small helper to query the V3D hardware version, since counter IDs
		and other details can differ between v4.2 and v7.1.

PATCH 5 improves the test coverage:

  - get-values-check-monotonically, which submits ten CL jobs in a row and
		asserts the counter values grow strictly between reads;
  - global-perfmon subtests covering input validation, single-perfmon
    semantics (EBUSY), implicit clear on perfmon destroy and on fd close,
		precedence over per-job perfmon (EAGAIN), and cross-fd accumulation;
  - perfmon-counter-isolation and perfmon-counter-isolation-subsequent-cl,
    which verify that the driver's queue serialization keeps pending CSD
    or earlier-CL work from leaking into a perfmon window.

Tested on Raspberry Pi 4 (V3D 4.2) and Raspberry Pi 5 (V3D 7.1).

Best regards,
- Maíra

Maíra Canal (5):
  tests/v3d_perfmon: Don't use the deprecated V3D_PERFCNT_NUM
  drm-uapi/v3d: Sync v3d UAPI
  lib/v3d: Allow callers to retrieve perfmon counter values
  lib/v3d: Add helper to query the V3D hardware version
  tests/v3d_perfmon: Add global perfmon and counter isolation tests

 include/drm-uapi/v3d_drm.h |  21 ++
 lib/igt_v3d.c              |  22 +-
 lib/igt_v3d.h              |   3 +-
 tests/v3d/v3d_perfmon.c    | 547 ++++++++++++++++++++++++++++++++++++-
 tests/v3d/v3d_submit_cl.c  |   4 +-
 tests/v3d/v3d_submit_csd.c |   4 +-
 6 files changed, 587 insertions(+), 14 deletions(-)

-- 
2.54.0


             reply	other threads:[~2026-05-08 12:46 UTC|newest]

Thread overview: 11+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-05-08 12:42 Maíra Canal [this message]
2026-05-08 12:42 ` [PATCH i-g-t 1/5] tests/v3d_perfmon: Don't use the deprecated V3D_PERFCNT_NUM Maíra Canal
2026-05-08 12:42 ` [PATCH i-g-t 2/5] drm-uapi/v3d: Sync v3d UAPI Maíra Canal
2026-05-08 12:42 ` [PATCH i-g-t 3/5] lib/v3d: Allow callers to retrieve perfmon counter values Maíra Canal
2026-05-08 12:42 ` [PATCH i-g-t 4/5] lib/v3d: Add helper to query the V3D hardware version Maíra Canal
2026-05-08 12:42 ` [PATCH i-g-t 5/5] tests/v3d_perfmon: Add global perfmon and counter isolation tests Maíra Canal
2026-05-12  6:50   ` Iago Toral
2026-05-09  0:21 ` ✓ i915.CI.BAT: success for tests/v3d_perfmon: Fix V3D 7.1 regression and extend perfmon testing Patchwork
2026-05-09  0:28 ` ✓ Xe.CI.BAT: " Patchwork
2026-05-09 13:32 ` ✗ Xe.CI.FULL: failure " Patchwork
2026-05-10  1:38 ` ✓ i915.CI.Full: success " Patchwork

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=20260508124446.1260672-2-mcanal@igalia.com \
    --to=mcanal@igalia.com \
    --cc=ashutosh.dixit@intel.com \
    --cc=bhanuprakash.modem@gmail.com \
    --cc=igt-dev@lists.freedesktop.org \
    --cc=itoral@igalia.com \
    --cc=juhapekka.heikkila@gmail.com \
    --cc=kamil.konieczny@linux.intel.com \
    --cc=karthik.b.s@intel.com \
    --cc=kernel-dev@igalia.com \
    --cc=mwen@igalia.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