All of lore.kernel.org
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Kathiravan Thirumoorthy" <kathiravan.thirumoorthy@oss.qualcomm.com>
Cc: devicetree@vger.kernel.org, robh@kernel.org, conor+dt@kernel.org
Subject: Re: [PATCH 3/4] arm64: dts: qcom: ipq9650: add and enable various device nodes
Date: Fri, 20 Nov 2026 09:32:18 +0000	[thread overview]
Message-ID: <sashiko-outbox-224853@kernel.org> (raw)
In-Reply-To: <20261120-ipq9650_various_devices-v1-3-14f53b84c42c@oss.qualcomm.com>

Thank you for your contribution! Sashiko AI review found 8 potential
issues to consider:
- [High] arm64: dts: qcom: ipq9650 QUPv3 I2C/SPI nodes omit
  assigned-clock-rates
- [Medium] arm64: dts: qcom: ipq9650 ssphy_0 reg size 0xa00 truncates
  0xe00 pcs_usb
- [Medium] arm64: dts: qcom: ipq9650 usb3 swaps dp_hs_phy_irq and
  dm_hs_phy_irq IRQs
- [Medium] arm64: dts: qcom: ipq9650 300 MHz OPP table attached to 200
  MHz sdhc
- [Low] arm64: dts: qcom: ipq9650 missing #address-cells/#size-cells on
  spi1 and i2c
- [Low] arm64: dts: qcom: ipq9650 sdhc_ice enabled in both common dtsi
  and board dts
- [Low] arm64: dts: qcom: ipq9650 qpic_bam omits status disabled in SoC
  dtsi
- [Low] arm64: dts: qcom: ipq9650 out-of-order DT nodes in soc@0 and
  rdp488-common
--

commit 72adc63cd895671ec07ed6eae7e42757d9770778
Author: Kathiravan Thirumoorthy <kathiravan.thirumoorthy@oss.qualcomm.com>

arm64: dts: qcom: ipq9650: add and enable various device nodes

This commit adds SoC device tree nodes for QUPv3 SPI and I2C, TCSR download
mode, QPIC SPI NAND and BAM, CPU idle states, USB2/USB3 and PHYs, DSU PMU,
LLCC, SMMU, and eMMC ICE on IPQ9650, and enables several peripherals on
RDP488.

> diff --git a/arch/arm64/boot/dts/qcom/ipq9650-rdp488-common.dtsi b/arch/arm64/boot/dts/qcom/ipq9650-rdp488-common.dtsi
> index e51d75bb9576e..bc2caf42faf5d 100644
> --- a/arch/arm64/boot/dts/qcom/ipq9650-rdp488-common.dtsi
> +++ b/arch/arm64/boot/dts/qcom/ipq9650-rdp488-common.dtsi
[ ... ]
> @@ -73,11 +154,44 @@ &sdhc {
>  	pinctrl-names = "default";
>  };
>  
> +&sdhc_ice {
> +	status = "okay";
> +};

[Severity: Low]
Should sdhc_ice only be enabled in ipq9650-rdp488.dts alongside sdhc rather
than in ipq9650-rdp488-common.dtsi?

ipq9650-rdp488.dts already enables sdhc_ice, so enabling it here duplicates
that override on the eMMC board and also probes the Inline Crypto Engine
(keeping GCC_SDCC1_ICE_CORE_CLK and GCC_SDCC1_AHB_CLK enabled) on NAND board
variants where sdhc is disabled.

> diff --git a/arch/arm64/boot/dts/qcom/ipq9650.dtsi b/arch/arm64/boot/dts/qcom/ipq9650.dtsi
> index 94269c9c8d6fe..a4091fd0c1eae 100644
> --- a/arch/arm64/boot/dts/qcom/ipq9650.dtsi
> +++ b/arch/arm64/boot/dts/qcom/ipq9650.dtsi
[ ... ]
> @@ -210,6 +270,61 @@ refgen_pcie: regulator@6c000 {
[ ... ]
> +		ssphy_0: phy@7d000 {
> +			compatible = "qcom,ipq9650-qmp-usb3-phy";
> +			reg = <0x0 0x0007d000 0x0 0xa00>;

[Severity: Medium]
Does the 0xa00 reg size on ssphy_0 truncate the PCS_USB register block at
offset 0xe00?

In drivers/phy/qualcomm/phy-qcom-qmp-usb.c, ipq9650_usb3phy_cfg uses
qmp_usb_offsets_v4, which sets .pcs_usb = 0x0e00, and
ipq9650_usb3_pcs_usb_tbl programs registers up to offset 0x0e18.  Because
qmp_usb_parse_dt() maps resource 0 and computes qmp->pcs_usb = base + 0x0e00,
qmp_usb_power_on() and runtime PM transitions access 0x400 bytes past the
declared 0xa00 resource range.  Should the reg size be 0x1000?

[ ... ]

> @@ -399,6 +519,84 @@ qupv3: geniqup@1ac0000 {
>  			#size-cells = <2>;
>  			ranges;
>  
> +			spi0: spi@1a80000 {
> +				compatible = "qcom,geni-spi";
> +				reg = <0x0 0x01a80000 0x0 0x4000>;
> +				clocks = <&gcc GCC_QUPV3_WRAP_SE0_CLK>;
> +				clock-names = "se";
> +				interrupts = <GIC_SPI 339 IRQ_TYPE_LEVEL_HIGH 0>;
> +				#address-cells = <1>;
> +				#size-cells = <0>;
> +				assigned-clocks = <&gcc GCC_QUPV3_WRAP_SE0_CLK>;
> +				operating-points-v2 = <&spi_opp_table_50mhz>;
> +
> +				status = "disabled";
> +			};
> +
> +			i2c0: i2c@1a84000 {
> +				compatible = "qcom,geni-i2c";
> +				reg = <0x0 0x01a84000 0x0 0x4000>;
> +				clocks = <&gcc GCC_QUPV3_WRAP_SE1_CLK>;
> +				clock-names = "se";
> +				clock-frequency = <400000>;
> +				interrupts = <GIC_SPI 340 IRQ_TYPE_LEVEL_HIGH 0>;
> +				assigned-clocks = <&gcc GCC_QUPV3_WRAP_SE1_CLK>;
> +				operating-points-v2 = <&i2c_opp_table_32mhz>;
> +
> +				status = "disabled";
> +			};

[Severity: High]
Are spi0, spi1, and i2c0..i2c4 missing assigned-clock-rates to match their
assigned-clocks properties?

Without assigned-clock-rates, of_clk_set_defaults() leaves the serial engine
clocks at their boot rate.  Unlike spi-geni-qcom, i2c-qcom-geni never calls
dev_pm_opp_set_rate() to apply i2c_opp_table_32mhz, and instead checks the
current rate of se.clk at probe time:

drivers/i2c/busses/i2c-qcom-geni.c:geni_i2c_clk_map_idx() {
	if (clk_get_rate(gi2c->se.clk) == GENI_SE_CLK_32MHZ) {
		req_freq = GENI_SE_CLK_32MHZ;
		itr = geni_i2c_clk_map_32mhz;
	} else {
		req_freq = GENI_SE_CLK_19P2MHZ;
		itr = geni_i2c_clk_map_19p2mhz;
	}
}

In drivers/clk/qcom/gcc-ipq9650.c, ftbl_gcc_qupv3_wrap_se0_clk_src and
ftbl_gcc_qupv3_wrap_se4_clk_src are derived from a 24 MHz XO reference and
support 32 MHz, not 19.2 MHz.  If GCC_QUPV3_WRAP_SE*_CLK is at its 24 MHz XO
default rather than pre-programmed to 32 MHz by firmware,
geni_se_clk_freq_match() fails with -EINVAL and i2c2 fails to probe on
RDP488.

[ ... ]

> +			spi1: spi@1a94000 {
> +				compatible = "qcom,geni-spi";
> +				reg = <0x0 0x01a94000 0x0 0x4000>;
> +				clocks = <&gcc GCC_QUPV3_WRAP_SE5_CLK>;
> +				clock-names = "se";
> +				interrupts = <GIC_SPI 344 IRQ_TYPE_LEVEL_HIGH 0>;
> +				assigned-clocks = <&gcc GCC_QUPV3_WRAP_SE5_CLK>;
> +				operating-points-v2 = <&spi_opp_table_50mhz>;
> +
> +				status = "disabled";
> +			};

[Severity: Low]
Should spi1 and i2c0..i2c4 define #address-cells = <1> and #size-cells = <0>
like spi0?  Without them, DT schema validation fails and any child SPI or I2C
devices added under those controllers inherit #address-cells = <2> and
#size-cells = <1>.

Conversely, the flash@0 nodes under &qpic_nand and &spi0 in
ipq9650-rdp488-common.dtsi set #address-cells = <1> and #size-cells = <1>
with no child subnodes, which Documentation/devicetree/bindings/mtd/mtd.yaml
marks as deprecated.

[ ... ]

> @@ -408,6 +606,75 @@ uart1: serial@1a98000 {
[ ... ]
> +		system-cache-controller@4000000 {
> +			compatible = "qcom,ipq9650-llcc";
> +			reg = <0x0 0x04000000 0x0 0x200000>;
> +			reg-names = "llcc0_base";
> +			interrupts = <GIC_SPI 68 IRQ_TYPE_LEVEL_HIGH 0>;
> +		};
> +
> +		usb2: usb@1e00000 {

[Severity: Low]
This isn't a bug, but system-cache-controller@4000000 is placed before
usb@1e00000 under soc@0 instead of in unit-address order.  Similarly, in
ipq9650-rdp488-common.dtsi, &qusb_phy_0 and &qusb_phy_1 are placed before
&i2c2, &spi0 before &sleep_clk, and &usb2 and &usb3 before &uart1 instead of
in alphabetical order.

[ ... ]

> @@ -429,6 +696,102 @@ sdhc: mmc@7804000 {
>  				      "core",
>  				      "xo";
>  			non-removable;
> +			qcom,ice = <&sdhc_ice>;
> +			operating-points-v2 = <&sdhc_opp_table_300mhz>;
> +			supports-cqe;
> +
> +			status = "disabled";
> +		};
> +
> +		sdhc_ice: crypto@7808000 {
> +			compatible = "qcom,ipq9650-inline-crypto-engine",
> +				     "qcom,inline-crypto-engine";
> +			reg = <0x0 0x07808000 0x0 0x2000>;
> +
> +			clocks = <&gcc GCC_SDCC1_ICE_CORE_CLK>,
> +				 <&gcc GCC_SDCC1_AHB_CLK>;
> +			clock-names = "core",
> +				      "iface";
> +
> +			status = "disabled";
> +		};

[Severity: Medium]
Is sdhc_opp_table_300mhz (300 MHz) attached to sdhc instead of configuring
the 300 MHz core clock on sdhc_ice?

sdhci_msm_probe() binds sdhc's operating-points-v2 table to its "core" clock
(GCC_SDCC1_APPS_CLK), but in drivers/clk/qcom/gcc-ipq9650.c,
ftbl_gcc_sdcc1_apps_clk_src only supports up to 200 MHz (and
ipq9650-rdp488-common.dtsi sets max-frequency = <192000000>).  300 MHz is the
rate in ftbl_gcc_sdcc1_ice_core_clk_src for sdhc_ice's GCC_SDCC1_ICE_CORE_CLK,
which currently has no rate configured in DT or drivers/soc/qcom/ice.c.
Also, Documentation/devicetree/bindings/mmc/qcom,sdhci-msm.yaml requires
required-opps on every OPP entry in an sdhci-msm opp-table.

> +
> +		qpic_bam: dma-controller@7984000 {
> +			compatible = "qcom,bam-v1.7.4", "qcom,bam-v1.7.0";
> +			reg = <0x0 0x7984000 0x0 0x1c000>;
> +			interrupts = <GIC_SPI 109 IRQ_TYPE_LEVEL_HIGH 0>;
> +			clocks = <&gcc GCC_QPIC_AHB_CLK>;
> +			clock-names = "bam_clk";
> +			#dma-cells = <1>;
> +			qcom,ee = <0>;
> +		};

[Severity: Low]
Should qpic_bam have status = "disabled" in ipq9650.dtsi like its consumer
qpic_nand?  Leaving it enabled causes bam_dma_probe() to probe the QPIC BAM
controller, keep GCC_QPIC_AHB_CLK enabled, and claim IRQ 109 on eMMC board
variants where qpic_nand is disabled.

[ ... ]

> +		usb3: usb@8a00000 {
[ ... ]
> +			interrupts-extended = <&intc GIC_SPI 409 IRQ_TYPE_LEVEL_HIGH 0>,
> +					      <&intc GIC_SPI 412 IRQ_TYPE_LEVEL_HIGH 0>,
> +					      <&intc GIC_SPI 413 IRQ_TYPE_LEVEL_HIGH 0>,
> +					      <&intc GIC_SPI 423 IRQ_TYPE_LEVEL_HIGH 0>,
> +					      <&intc GIC_SPI 424 IRQ_TYPE_LEVEL_HIGH 0>;
> +			interrupt-names = "dwc_usb3",
> +					  "pwr_event",
> +					  "qusb2_phy",
> +					  "dp_hs_phy_irq",
> +					  "dm_hs_phy_irq";

[Severity: Medium]
Are GIC_SPI 423 and GIC_SPI 424 swapped relative to "dp_hs_phy_irq" and
"dm_hs_phy_irq" on usb3?

On usb2 above (where GIC_SPI 388 is dp_hs_phy_irq and GIC_SPI 387 is
dm_hs_phy_irq) and on ipq5424.dtsi (where GIC_SPI 423 is dm_hs_phy_irq and
GIC_SPI 424 is dp_hs_phy_irq), the lower SPI number is DM and the higher SPI
number is DP.  With 423 mapped to dp_hs_phy_irq and 424 mapped to
dm_hs_phy_irq, dwc3_qcom_enable_port_interrupts() monitors D- instead of D+
for High-Speed/Full-Speed devices and D+ instead of D- for Low-Speed devices
during suspend, breaking USB2 wakeup and disconnect detection on usb3.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261120-ipq9650_various_devices-v1-0-14f53b84c42c@oss.qualcomm.com?part=3

  reply	other threads:[~2026-11-20  9:32 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-20  8:59 [PATCH 0/4] arm64: dts: qcom: ipq9650: add and enable various devices Kathiravan Thirumoorthy
2026-11-20  8:59 ` [PATCH 1/4] arm64: dts: qcom: ipq9650: fix the TCSR mutex address range Kathiravan Thirumoorthy
2026-11-20  8:59 ` [PATCH 2/4] arm64: dts: qcom: ipq9650: add RDP488 common DT file Kathiravan Thirumoorthy
2026-11-20  8:59 ` [PATCH 3/4] arm64: dts: qcom: ipq9650: add and enable various device nodes Kathiravan Thirumoorthy
2026-11-20  9:32   ` sashiko-bot [this message]
2026-11-20  8:59 ` [PATCH 4/4] arm64: dts: qcom: ipq9650: Add RDP488 NAND variant Kathiravan Thirumoorthy
2026-11-20  9:07   ` sashiko-bot

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=sashiko-outbox-224853@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=conor+dt@kernel.org \
    --cc=devicetree@vger.kernel.org \
    --cc=kathiravan.thirumoorthy@oss.qualcomm.com \
    --cc=robh@kernel.org \
    --cc=sashiko-reviews@lists.linux.dev \
    /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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.