From: Arun R Murthy <arun.r.murthy@intel.com>
To: intel-gfx@lists.freedesktop.org, intel-xe@lists.freedesktop.org
Cc: "Jani Nikula" <jani.nikula@linux.intel.com>,
"Arun R Murthy" <arun.r.murthy@intel.com>,
"Polo-François Poli" <polo.code.dev@gmail.com>
Subject: [PATCH 1/2] drm/i915/display: Seed skl_cdclk_init_hw() with SKU-max, not zero
Date: Tue, 04 Aug 2026 14:58:30 +0530 [thread overview]
Message-ID: <20260804-cdclk-v1-1-484864b1aafd@intel.com> (raw)
In-Reply-To: <20260804-cdclk-v1-0-484864b1aafd@intel.com>
skl_cdclk_init_hw() has a recovery path that runs when skl_sanitize_cdclk()
has decided the pre-OS CDCLK state is not usable and zeroes
display->cdclk.hw.
Seed with the SKU-permitted maximum instead (as read from SKL_DFSM), so
the recovery path lands on a cdclk that satisfies every pipe on the SKU.
The very first atomic commit that follows will dial cdclk down to what
active pipes actually need via the normal skl_modeset_calc_cdclk() path.
Closes: https://gitlab.freedesktop.org/drm/i915/kernel/-/work_items/16595
Signed-off-by: Arun R Murthy <arun.r.murthy@intel.com>
Tested-by: Polo-François Poli <polo.code.dev@gmail.com>
---
drivers/gpu/drm/i915/display/intel_cdclk.c | 65 +++++++++++++++++++++++++++++-
1 file changed, 64 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/i915/display/intel_cdclk.c b/drivers/gpu/drm/i915/display/intel_cdclk.c
index a53d887271778193fd400b9199ad841d5d9730fe..19f3ef8adfcded378428924539db9336834278ea 100644
--- a/drivers/gpu/drm/i915/display/intel_cdclk.c
+++ b/drivers/gpu/drm/i915/display/intel_cdclk.c
@@ -1288,6 +1288,31 @@ static void skl_sanitize_cdclk(struct intel_display *display)
display->cdclk.hw.vco = ~0;
}
+/*
+ * Seed cdclk from the SKU-permitted maximum. Used by skl_cdclk_init_hw()'s
+ * recovery path when skl_sanitize_cdclk() has decided the pre-OS CDCLK state
+ * is not usable and has zeroed display->cdclk.hw.
+ * Seed with the SKU top instead: it is guaranteed to satisfy every pipe on
+ * the platform, and the very first atomic commit will dial cdclk down via
+ * skl_modeset_calc_cdclk() once per-pipe min_cdclk is known.
+ */
+static int skl_cdclk_init_hw_seed(struct intel_display *display, int vco)
+{
+ u32 limit = intel_de_read(display, SKL_DFSM) & SKL_DFSM_CDCLK_LIMIT_MASK;
+ int max_cdclk;
+
+ if (limit == SKL_DFSM_CDCLK_LIMIT_675)
+ max_cdclk = 617143;
+ else if (limit == SKL_DFSM_CDCLK_LIMIT_540)
+ max_cdclk = 540000;
+ else if (limit == SKL_DFSM_CDCLK_LIMIT_450)
+ max_cdclk = 432000;
+ else
+ max_cdclk = 308571;
+
+ return skl_calc_cdclk(max_cdclk, vco);
+}
+
static void skl_cdclk_init_hw(struct intel_display *display)
{
struct intel_cdclk_config cdclk_config;
@@ -1311,9 +1336,14 @@ static void skl_cdclk_init_hw(struct intel_display *display)
cdclk_config.vco = display->cdclk.skl_preferred_vco_freq;
if (cdclk_config.vco == 0)
cdclk_config.vco = 8100000;
- cdclk_config.cdclk = skl_calc_cdclk(0, cdclk_config.vco);
+
+ cdclk_config.cdclk = skl_cdclk_init_hw_seed(display, cdclk_config.vco);
cdclk_config.voltage_level = skl_calc_voltage_level(cdclk_config.cdclk);
+ drm_dbg_kms(display->drm,
+ "cdclk: init seed cdclk=%d vco=%d volt=%d (SKU-max, no per-pipe min_cdclk yet)\n",
+ cdclk_config.cdclk, cdclk_config.vco, cdclk_config.voltage_level);
+
skl_set_cdclk(display, &cdclk_config, INVALID_PIPE);
}
@@ -4231,6 +4261,39 @@ void intel_cdclk_read_hw(struct intel_display *display)
cdclk_state->logical = display->cdclk.hw;
}
+/**
+ * intel_cdclk_sync_hw_state - Adopt the HW-asserted cdclk into SW state.
+ * @display: display instance
+ * @hw_config: the cdclk configuration as currently read back from HW
+ *
+ * This helper reconciles the SW state to match what HW is currently asserting.
+ * It updates display->cdclk.hw and the committed cdclk_state's actual/logical
+ * so that any following atomic commit sees a truthful baseline and will
+ * naturally attach a cdclk_state (via intel_cdclk_atomic_check()) to raise
+ * cdclk when the workload demands it.
+ */
+void intel_cdclk_sync_hw_state(struct intel_display *display,
+ const struct intel_cdclk_config *hw_config)
+{
+ struct intel_cdclk_state *cdclk_state =
+ to_intel_cdclk_state(display->cdclk.obj.state);
+
+ if (!intel_cdclk_clock_changed(&display->cdclk.hw, hw_config))
+ return;
+
+ drm_dbg_kms(display->drm,
+ "cdclk: syncing SW state to HW after drift (sw %d/vco %d/ref %d -> hw %d/vco %d/ref %d)\n",
+ display->cdclk.hw.cdclk, display->cdclk.hw.vco, display->cdclk.hw.ref,
+ hw_config->cdclk, hw_config->vco, hw_config->ref);
+
+ display->cdclk.hw.cdclk = hw_config->cdclk;
+ display->cdclk.hw.vco = hw_config->vco;
+ display->cdclk.hw.ref = hw_config->ref;
+
+ cdclk_state->actual = display->cdclk.hw;
+ cdclk_state->logical = display->cdclk.hw;
+}
+
static int calc_cdclk(const struct intel_crtc_state *crtc_state, int min_cdclk)
{
struct intel_display *display = to_intel_display(crtc_state);
--
2.25.1
next prev parent reply other threads:[~2026-08-04 9:29 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-08-04 9:28 [PATCH 0/2] cdclk mismatch with sw/hw Arun R Murthy
2026-08-04 9:28 ` Arun R Murthy [this message]
2026-08-04 9:28 ` [PATCH 2/2] drm/i915/display: Reconcile cdclk SW state with HW on DC-off exit Arun R Murthy
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=20260804-cdclk-v1-1-484864b1aafd@intel.com \
--to=arun.r.murthy@intel.com \
--cc=intel-gfx@lists.freedesktop.org \
--cc=intel-xe@lists.freedesktop.org \
--cc=jani.nikula@linux.intel.com \
--cc=polo.code.dev@gmail.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