rockchip: backport MMC CQE fixes

Added missing platform data struct for RK3576 to fix boot issues,
fixed general stability issues and also improved random write
performance.

Signed-off-by: Tianling Shen <cnsztl@immortalwrt.org>
Link: https://github.com/openwrt/openwrt/pull/21028
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
This commit is contained in:
Tianling Shen 2025-12-03 14:36:32 +08:00 committed by Hauke Mehrtens
parent 5aeaddc0c8
commit 4582d25a66
3 changed files with 141 additions and 0 deletions

View File

@ -0,0 +1,55 @@
From 69cc9d4075855661268327c38c9b0e71ac37eb1c Mon Sep 17 00:00:00 2001
From: Sebastian Reichel <sebastian.reichel@collabora.com>
Date: Fri, 21 Nov 2025 17:26:59 +0100
Subject: [PATCH] mmc: sdhci-of-dwcmshc: Fix command queue support for RK3576
When I added command queue engine (CQE) support for the Rockchip eMMC
controller, I missed that RK3576 has a separate platform data struct.
While things are working fine on RK3588 (I tested the ROCK 5B) and
the suspend issue is fixed on the RK3576 (I tested the Sige5), this
results in stability issues. By also adding the necessary hooks for
the RK3576 platform the following problems can be avoided:
[ 15.606895] mmc0: running CQE recovery
[ 15.616189] mmc0: running CQE recovery
[...]
[ 25.911484] mmc0: running CQE recovery
[ 25.926305] mmc0: running CQE recovery
[ 25.927468] mmc0: running CQE recovery
[...]
[ 26.255719] mmc0: running CQE recovery
[ 26.257162] ------------[ cut here ]------------
[ 26.257581] mmc0: cqhci: spurious TCN for tag 31
[ 26.258034] WARNING: CPU: 0 PID: 0 at drivers/mmc/host/cqhci-core.c:796 cqhci_irq+0x440/0x68c
[ 26.263786] CPU: 0 UID: 0 PID: 0 Comm: swapper/0 Not tainted 6.18.0-rc6-gd984ebbf0d15 #1 PREEMPT
[ 26.264561] Hardware name: ArmSoM Sige5 (DT)
[...]
[ 26.272748] Call trace:
[ 26.272964] cqhci_irq+0x440/0x68c (P)
[ 26.273296] dwcmshc_cqe_irq_handler+0x54/0x88
[ 26.273689] sdhci_irq+0xbc/0x1200
[ 26.273991] __handle_irq_event_percpu+0x54/0x1d0
[...]
Note that the above problems do not necessarily happen with every boot.
Reported-by: Adrian Hunter <adrian.hunter@intel.com>
Closes: https://lore.kernel.org/linux-rockchip/01949bc9-4873-498b-ac7d-f008393ccc4c@intel.com/
Fixes: fda1e0af7c28f ("mmc: sdhci-of-dwcmshc: Add command queue support for rockchip SOCs")
Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com>
Reviewed-by: Shawn Lin <shawn.lin@rock-chips.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
---
drivers/mmc/host/sdhci-of-dwcmshc.c | 1 +
1 file changed, 1 insertion(+)
--- a/drivers/mmc/host/sdhci-of-dwcmshc.c
+++ b/drivers/mmc/host/sdhci-of-dwcmshc.c
@@ -1333,6 +1333,7 @@ static const struct dwcmshc_pltfm_data s
.quirks2 = SDHCI_QUIRK2_PRESET_VALUE_BROKEN |
SDHCI_QUIRK2_CLOCK_DIV_ZERO_BROKEN,
},
+ .cqhci_host_ops = &rk35xx_cqhci_ops,
.init = dwcmshc_rk35xx_init,
.postinit = dwcmshc_rk3576_postinit,
};

View File

@ -0,0 +1,42 @@
From c7ce6453b769c45006ed4983762f81e130878171 Mon Sep 17 00:00:00 2001
From: Shawn Lin <shawn.lin@rock-chips.com>
Date: Wed, 26 Nov 2025 07:26:39 +0800
Subject: [PATCH] mmc: sdhci-of-dwcmshc: Disable internal clock auto gate for
Rockchip SOCs
Enabling CMDQ support can lead to random occurrences of the error log when
there are RPMB access and data flush executed:
"mmc2: Timeout waiting for hardware interrupt."
Enabling CMDQ and then issuing a DCMD as the final command before disabling
it causes the eMMC controller to auto-gate its internal clock. Chip simulation
shows this results in a state machine mismatch after CMDQ mode exit, triggering
data-timeout errors for all subsequent read and write operations.
Therefore, the auto-clock-gate function must be disabled whenever CMDQ is
enabled.
Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Fixes: fda1e0af7c28 ("mmc: sdhci-of-dwcmshc: Add command queue support for rockchip SOCs")
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
---
drivers/mmc/host/sdhci-of-dwcmshc.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
--- a/drivers/mmc/host/sdhci-of-dwcmshc.c
+++ b/drivers/mmc/host/sdhci-of-dwcmshc.c
@@ -696,10 +696,11 @@ static void dwcmshc_rk3568_set_clock(str
sdhci_set_clock(host, clock);
- /* Disable cmd conflict check */
+ /* Disable cmd conflict check and internal clock gate */
reg = dwc_priv->vendor_specific_area1 + DWCMSHC_HOST_CTRL3;
extra = sdhci_readl(host, reg);
extra &= ~BIT(0);
+ extra |= BIT(4);
sdhci_writel(host, extra, reg);
if (clock <= 52000000) {

View File

@ -0,0 +1,44 @@
From 79cf71c0b177c0e23d411e2469435e2c2f83f563 Mon Sep 17 00:00:00 2001
From: Shawn Lin <shawn.lin@rock-chips.com>
Date: Wed, 26 Nov 2025 07:26:40 +0800
Subject: [PATCH] mmc: sdhci-of-dwcmshc: reduce CIT for better performance
CQHCI_SSC1.CIT indicates to the CQE the polling period to use for
periodic SEND_QUEUE_STATUS (CMD13) polling. Some eMMCs have only one
hardware queue, and CMD13 can only query one slot at a time for data
transmission, which cannot be processed in parallel. Modifying the
CMD13 query interval can increase the query frequency and improve
random write performance.
Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
---
drivers/mmc/host/cqhci.h | 1 +
drivers/mmc/host/sdhci-of-dwcmshc.c | 5 +++++
2 files changed, 6 insertions(+)
--- a/drivers/mmc/host/cqhci.h
+++ b/drivers/mmc/host/cqhci.h
@@ -93,6 +93,7 @@
/* send status config 1 */
#define CQHCI_SSC1 0x40
#define CQHCI_SSC1_CBC_MASK GENMASK(19, 16)
+#define CQHCI_SSC1_CIT_MASK GENMASK(15, 0)
/* send status config 2 */
#define CQHCI_SSC2 0x44
--- a/drivers/mmc/host/sdhci-of-dwcmshc.c
+++ b/drivers/mmc/host/sdhci-of-dwcmshc.c
@@ -614,6 +614,11 @@ static void rk35xx_sdhci_cqe_pre_enable(
struct dwcmshc_priv *dwc_priv = sdhci_pltfm_priv(pltfm_host);
u32 reg;
+ /* Set Send Status Command Idle Timer to 10.66us (256 * 1 / 24) */
+ reg = sdhci_readl(host, dwc_priv->vendor_specific_area2 + CQHCI_SSC1);
+ reg = (reg & ~CQHCI_SSC1_CIT_MASK) | 0x0100;
+ sdhci_writel(host, reg, dwc_priv->vendor_specific_area2 + CQHCI_SSC1);
+
reg = sdhci_readl(host, dwc_priv->vendor_specific_area2 + CQHCI_CFG);
reg |= CQHCI_ENABLE;
sdhci_writel(host, reg, dwc_priv->vendor_specific_area2 + CQHCI_CFG);