[PATCH v4 0/3] mmc: sdhci-cadence: SD6HC DDR50/DDR52 read-path tuning and fixes

From: tze . yee . ng

Date: Thu Nov 19 2026 - 04:46:11 EST


From: Tze Yee Ng <tze.yee.ng@xxxxxxxxxx>

This series enables the extended-read DDR modes - SD DDR50 and eMMC
DDR52 - on the Cadence SD6HC (SD/eMMC v6) PHY. Neither mode runs a host
tuning sequence (CMD19/CMD21), so the read path must be centred by
static, board/SoC-characterised PHY settings. Bringing it up on Agilex5
also exposed PHY issues affecting the existing tuned modes.

- Patch 1: abort a failed tuning command on SD6HC. The software tuning
loop never aborted a command that errored, so after a CMD21 command
error the eMMC could stay in the Data state and time out the CMD6
that HS400 selection issues next; mmc_send_abort_tuning() clears it
(a no-op for SD). Limited to SD6HC since the path is shared with the
untested SD4HC.
- Patch 2: document two optional SD6HC read-path properties for the
extended-read DDR modes (SD DDR50, eMMC DDR52).
- Patch 3: program PHONY_DQS_TIMING, which was left 0 in all modes; in
extended-read mode it must be the rebar pulse width in clk_phy
cycles minus 1, else one DDR beat is duplicated. The width is now
derived from the SDCLK/sdmclk divider rather than hard-coded. The
same patch reads the new properties at probe and applies them in
DDR50 and DDR52, and makes the DQS_TIMING update clear USE_LPBK_DQS
before reprogramming it so the loopback/phony selection takes
effect.

The new cdns,ddr-* properties are unused in this series; they will be
consumed by the socfpga_agilex5_socdk and socfpga_agilex5_socdk_emmc
device trees, submitted as a separate series.

Tested on Agilex5 SoCDK: SD DDR50, eMMC DDR52 and eMMC HS400.

Signed-off-by: Tze Yee Ng <tze.yee.ng@xxxxxxxxxx>
---
Changes in v4:
- Merged v3 2/4 (PHONY_DQS_TIMING) and v3 4/4 (read DDR tuning from
device tree) into a single driver patch, now 3/3, and moved it after
the binding patch. The series is now 3 patches. (Tanmay Kathpalia)
- Patch 1: limited the abort to SD6HC (host->version >= SDHCI_SPEC_420)
because sdhci_cdns_execute_tuning() is shared with SD4HC, which has
not been tested; reworded the commit message accordingly. Added Tanmay's
Reviewed-by.
- Patch 2: dropped the cdns,ddr-phony-dqs-timing property; the phony DQS
timing is always derived by the driver. (Tanmay Kathpalia)
- Patch 2: dropped Krzysztof's Reviewed-by since property set changed. The
two remaining properties are unchanged from v3; Krzysztof, happy to
carry your tag again if the reduced binding looks fine.
- Patch 3: derive the PHONY_DQS_TIMING rebar pulse width from the
SDCLK/sdmclk divider (t_sdclk / (2 * t_sdmclk) - 1, clamped to the 6-bit
field) instead of the hard-coded 2 clk_phy cycles, and apply it in every
extended-read mode instead of only DDR50/DDR52. Dropped the
cdns,ddr-phony-dqs-timing override. Stored ddr_read_dqs_delay as u8.
- Patch 3: dropped Adrian's Acked-by from v3 2/4 since the patch was
merged and substantially reworked; Adrian, please re-review.

Changes in v3:
- Patch 1: replaced the post-tuning PHY settle (v2 1/4) with aborting
the failed tuning command. Added Suggested-by. (Tanmay Kathpalia)
- Patch 2: unchanged; retains Adrian's Acked-by.
- Patch 3: re-expressed cdns,ddr-read-dqs-delay-ps as
cdns,ddr-read-dqs-delay - a count of DLL delay elements (0-255, each
cdns,delay-element-ps wide) that maps 1:1 to the 8-bit PHY field -
instead of a picosecond value; added $ref uint32 and maximum: 255,
dropped the redundant minimum: 0, and used plain description: text.
(Krzysztof Kozlowski)
- Patch 4: read cdns,ddr-read-dqs-delay directly as the delay-element
count and write it to the 8-bit field, dropping the v2 ps->element
conversion (divide by cdns,delay-element-ps) and the FIELD_MAX()
clamp. Hoisted the repeated DDR-mode test to a local, removed the
probe-time firmware-value warning, and unwrapped lines to 100 cols.
(Adrian Hunter, Krzysztof Kozlowski)

- Link to v3: https://lore.kernel.org/all/20261106-sd_cadence_ddr50_v3-v3-0-cb50afece614@xxxxxxxxxx/
- Link to v2: https://lore.kernel.org/all/cover.1791459448.git.tze.yee.ng@xxxxxxxxxx/
- Link to v1: https://lore.kernel.org/all/cover.1790074790.git.tze.yee.ng@xxxxxxxxxx/
---
To: Adrian Hunter <adrian.hunter@xxxxxxxxx>
To: Tanmay Kathpalia <tanmay.kathpalia@xxxxxxxxxx>
To: Ulf Hansson <ulfh@xxxxxxxxxx>
To: Rob Herring <robh@xxxxxxxxxx>
To: Krzysztof Kozlowski <krzk+dt@xxxxxxxxxx>
To: Conor Dooley <conor+dt@xxxxxxxxxx>
Cc: linux-mmc@xxxxxxxxxxxxxxx
Cc: linux-kernel@xxxxxxxxxxxxxxx
Cc: devicetree@xxxxxxxxxxxxxxx

---
Tze Yee Ng (3):
mmc: sdhci-cadence: abort failed tuning command on SD6HC
dt-bindings: mmc: cdns,sd6hc: add SD6HC DDR50 read-path tuning
mmc: sdhci-cadence6: tune SD6HC DDR50/DDR52 read path

.../devicetree/bindings/mmc/cdns,sd6hc.yaml | 19 ++++++++
drivers/mmc/host/sdhci-cadence-core.c | 12 ++++-
drivers/mmc/host/sdhci-cadence-phy-v6.c | 52 +++++++++++++++++++++-
3 files changed, 81 insertions(+), 2 deletions(-)
---
base-commit: 2febfc8acbf712439b6d30c99b108641d881fd75
change-id: 20261106-sd_cadence_ddr50_v3-35f970f31364

Best regards,
--
Tze Yee Ng <tze.yee.ng@xxxxxxxxxx>