* [PATCH v6 0/2] usb: cdns3: support configurations without DRD block
@ 2026-05-13 11:44 Pawel Laszczak via B4 Relay
2026-05-13 11:44 ` [PATCH v6 1/2] dt-bindings: usb: cdns3: Add no_drd property Pawel Laszczak via B4 Relay
2026-05-13 11:44 ` [PATCH v6 2/2] usb: cdnsp: Add support for device-only configuration Pawel Laszczak via B4 Relay
0 siblings, 2 replies; 6+ messages in thread
From: Pawel Laszczak via B4 Relay @ 2026-05-13 11:44 UTC (permalink / raw)
To: Peter Chen, Roger Quadros, Greg Kroah-Hartman, Rob Herring,
Krzysztof Kozlowski, Conor Dooley
Cc: linux-usb, devicetree, linux-kernel, Pawel Laszczak,
Bjorn Helgaas
usb: cdns3: support configurations without DRD block
This series adds support for Cadence USBSSP controllers in hardware
layouts where the Dual-Role Device (DRD) register block is either
missing or inaccessible.
In such configurations, the controller is hardwired to a single role
(either host or device) and the driver must skip all OTG/DRD register
accesses to avoid bus errors or incorrect role detection.
The solution introduces a new 'no_drd' property that can be passed
via DT or software nodes. When set, the driver:
1. Skips DRD register mapping and IRQ requests.
2. Uses a different BAR indexing logic for PCI-based configurations
(32-bit addressing layout).
3. Hardwires the role based on 'dr_mode'.
Signed-off-by: Pawel Laszczak <pawell@cadence.com>
---
Note: This series is based on current linux-next. I am aware of Peter
Chen's recent refactoring series ("usb: cdns3: plat: Expose platform
core driver as library"). Although there is a minor conflict in
cdns3-plat.c, Peter has already provided an Acked-by for this version.
I am happy to provide a rebased v7 as soon as Peter's changes land in
linux-next if required.
---
v6:
- Fixed validation error for 'interrupt-names' by correcting
the items definition.
- Adjusted 'minItems'/'maxItems' to properly support the optional
'wakeup' interrupt.
- Fixed 'too long' schema error in examples.
v5:
- Implemented strict conditional validation using if-then-else logic.
- Enforced 2 register/interrupt items and required 'dr_mode'
(host or peripheral) when 'no_drd' is present.
- Enforced the standard 3 register/interrupt items (otg, host, dev)
when 'no_drd' is absent to ensure backward compatibility.
- Updated 'reg-names' and 'interrupt-names' to use enums in the main
properties section to support flexible resource ordering during
validation.
v4:
- Added DT binding documentation for the 'no_drd' property.
- Relaxed 'reg' and 'interrupts' requirements in the DT schema (minItems 2)
to allow configurations where the OTG/DRD register block is missing.
- Moved PCI_DEVICE_ID_CDNS_UDC_USBSSP from pci_ids.h to cdnsp-pci.c
to keep the global PCI ID list clean.
v3:
- Improved descriptions and comments for better clarity.
- Introduced the 'no_drd' property to indicate missing DRD register block.
- Added support for fixed host-only and device-only configurations.
- Ensured cdns_otg_disable_irq is called only when no_drd is false.
- Updated cdns_drd_gadget_on/off to ensure PHY mode is correctly
handled even if DRD is disabled.
v2:
- Changed otg_irq to be optional.
- Added cdns->no_drd check in cdns_power_is_lost.
- Added cdns->no_drd check in cdns_get_id.
---
Pawel Laszczak (2):
dt-bindings: usb: cdns3: Add no_drd property
usb: cdnsp: Add support for device-only configuration
.../devicetree/bindings/usb/cdns,usb3.yaml | 70 +++++++++++++++++++---
drivers/usb/cdns3/cdns3-plat.c | 26 ++++----
drivers/usb/cdns3/cdnsp-pci.c | 47 ++++++++++++---
drivers/usb/cdns3/core.c | 3 +-
drivers/usb/cdns3/core.h | 4 ++
drivers/usb/cdns3/drd.c | 44 +++++++++++++-
6 files changed, 162 insertions(+), 32 deletions(-)
---
base-commit: e98d21c170b01ddef366f023bbfcf6b31509fa83
change-id: 20260512-b4-no_drd_config-f530f1f16d8a
Best regards,
--
Pawel Laszczak <pawell@cadence.com>
^ permalink raw reply [flat|nested] 6+ messages in thread* [PATCH v6 1/2] dt-bindings: usb: cdns3: Add no_drd property 2026-05-13 11:44 [PATCH v6 0/2] usb: cdns3: support configurations without DRD block Pawel Laszczak via B4 Relay @ 2026-05-13 11:44 ` Pawel Laszczak via B4 Relay 2026-05-13 18:59 ` Conor Dooley 2026-05-14 4:27 ` sashiko-bot 2026-05-13 11:44 ` [PATCH v6 2/2] usb: cdnsp: Add support for device-only configuration Pawel Laszczak via B4 Relay 1 sibling, 2 replies; 6+ messages in thread From: Pawel Laszczak via B4 Relay @ 2026-05-13 11:44 UTC (permalink / raw) To: Peter Chen, Roger Quadros, Greg Kroah-Hartman, Rob Herring, Krzysztof Kozlowski, Conor Dooley Cc: linux-usb, devicetree, linux-kernel, Pawel Laszczak From: Pawel Laszczak <pawell@cadence.com> Introduce a new boolean property 'no_drd' for Cadence USBSS/USBSSP controllers to support hardware configurations where the Dual-Role Device (DRD) register block is missing or inaccessible. When 'no_drd' is present: - The 'otg' register and interrupt resources are not required. - The 'reg' and 'interrupts' properties are restricted to 2 items (host and device). - 'dr_mode' must be explicitly set to either 'host' or 'peripheral'. When 'no_drd' is absent, the binding maintains backward compatibility by requiring all 3 resource sets (otg, host, dev). To achieve this, the schema is updated with an if-then-else logic and 'reg-names'/'interrupt-names' use enums to allow flexible ordering during validation. Signed-off-by: Pawel Laszczak <pawell@cadence.com> --- v6: - Fixed validation error for 'interrupt-names' by correcting the items definition. - Adjusted 'minItems'/'maxItems' to properly support the optional 'wakeup' interrupt. - Fixed 'too long' schema error in examples. v5: - Implemented strict conditional validation using if-then-else logic. - Enforced 2 register/interrupt items and required 'dr_mode' (host or peripheral) when 'no_drd' is present. - Enforced the standard 3 register/interrupt items (otg, host, dev) when 'no_drd' is absent to ensure backward compatibility. - Updated 'reg-names' and 'interrupt-names' to use enums in the main properties section to support flexible resource ordering during validation. --- --- .../devicetree/bindings/usb/cdns,usb3.yaml | 70 +++++++++++++++++++--- 1 file changed, 61 insertions(+), 9 deletions(-) diff --git a/Documentation/devicetree/bindings/usb/cdns,usb3.yaml b/Documentation/devicetree/bindings/usb/cdns,usb3.yaml index 2d95fb7321af..717892a05dcd 100644 --- a/Documentation/devicetree/bindings/usb/cdns,usb3.yaml +++ b/Documentation/devicetree/bindings/usb/cdns,usb3.yaml @@ -20,19 +20,21 @@ properties: const: cdns,usb3 reg: + minItems: 2 items: - description: OTG controller registers - description: XHCI Host controller registers - description: DEVICE controller registers reg-names: + minItems: 2 items: - - const: otg - - const: xhci - - const: dev + - enum: [ otg, xhci, dev ] + - enum: [ otg, xhci, dev ] + - enum: [ otg, xhci, dev ] interrupts: - minItems: 3 + minItems: 2 items: - description: XHCI host controller interrupt - description: Device controller interrupt @@ -41,12 +43,12 @@ properties: cleared by xhci core, this interrupt is optional interrupt-names: - minItems: 3 + minItems: 2 items: - - const: host - - const: peripheral - - const: otg - - const: wakeup + - enum: [ host, peripheral, otg, wakeup ] + - enum: [ host, peripheral, otg, wakeup ] + - enum: [ host, peripheral, otg, wakeup ] + - enum: [ host, peripheral, otg, wakeup ] port: $ref: /schemas/graph.yaml#/properties/port @@ -79,6 +81,13 @@ properties: description: Enable resetting of PHY if Rx fail is detected type: boolean + no_drd: + description: + Indicates that the Dual-Role Device (DRD) register block is not + implemented or is inaccessible. In this case, the controller + must operate in a fixed peripheral or host mode. + type: boolean + dependencies: port: [ usb-role-switch ] @@ -93,6 +102,49 @@ allOf: - $ref: usb-drd.yaml# - $ref: usb-xhci.yaml# + - if: + properties: + no_drd: true + required: + - no_drd + then: + required: + - dr_mode + properties: + reg: + maxItems: 2 + reg-names: + items: + - const: xhci + - const: dev + interrupts: + maxItems: 2 + interrupt-names: + items: + - const: host + - const: peripheral + dr_mode: + enum: [host, peripheral] + else: + properties: + reg: + minItems: 3 + reg-names: + items: + - const: otg + - const: xhci + - const: dev + interrupts: + minItems: 3 + maxItems: 4 + interrupt-names: + minItems: 3 + items: + - const: host + - const: peripheral + - const: otg + - const: wakeup + unevaluatedProperties: false examples: -- 2.43.0 ^ permalink raw reply related [flat|nested] 6+ messages in thread
* Re: [PATCH v6 1/2] dt-bindings: usb: cdns3: Add no_drd property 2026-05-13 11:44 ` [PATCH v6 1/2] dt-bindings: usb: cdns3: Add no_drd property Pawel Laszczak via B4 Relay @ 2026-05-13 18:59 ` Conor Dooley 2026-05-14 4:27 ` sashiko-bot 1 sibling, 0 replies; 6+ messages in thread From: Conor Dooley @ 2026-05-13 18:59 UTC (permalink / raw) To: pawell Cc: Peter Chen, Roger Quadros, Greg Kroah-Hartman, Rob Herring, Krzysztof Kozlowski, Conor Dooley, linux-usb, devicetree, linux-kernel [-- Attachment #1: Type: text/plain, Size: 5771 bytes --] On Wed, May 13, 2026 at 01:44:19PM +0200, Pawel Laszczak via B4 Relay wrote: > From: Pawel Laszczak <pawell@cadence.com> > > Introduce a new boolean property 'no_drd' for Cadence USBSS/USBSSP > controllers to support hardware configurations where the Dual-Role > Device (DRD) register block is missing or inaccessible. > > When 'no_drd' is present: > - The 'otg' register and interrupt resources are not required. > - The 'reg' and 'interrupts' properties are restricted to 2 items > (host and device). > - 'dr_mode' must be explicitly set to either 'host' or 'peripheral'. > > When 'no_drd' is absent, the binding maintains backward compatibility > by requiring all 3 resource sets (otg, host, dev). > > To achieve this, the schema is updated with an if-then-else logic > and 'reg-names'/'interrupt-names' use enums to allow flexible > ordering during validation. > > Signed-off-by: Pawel Laszczak <pawell@cadence.com> > --- > v6: > - Fixed validation error for 'interrupt-names' by correcting > the items definition. > - Adjusted 'minItems'/'maxItems' to properly support the optional > 'wakeup' interrupt. > - Fixed 'too long' schema error in examples. > > v5: > - Implemented strict conditional validation using if-then-else logic. > - Enforced 2 register/interrupt items and required 'dr_mode' > (host or peripheral) when 'no_drd' is present. > - Enforced the standard 3 register/interrupt items (otg, host, dev) > when 'no_drd' is absent to ensure backward compatibility. > - Updated 'reg-names' and 'interrupt-names' to use enums in the main > properties section to support flexible resource ordering during > validation. > --- > --- > .../devicetree/bindings/usb/cdns,usb3.yaml | 70 +++++++++++++++++++--- > 1 file changed, 61 insertions(+), 9 deletions(-) > > diff --git a/Documentation/devicetree/bindings/usb/cdns,usb3.yaml b/Documentation/devicetree/bindings/usb/cdns,usb3.yaml > index 2d95fb7321af..717892a05dcd 100644 > --- a/Documentation/devicetree/bindings/usb/cdns,usb3.yaml > +++ b/Documentation/devicetree/bindings/usb/cdns,usb3.yaml > @@ -20,19 +20,21 @@ properties: > const: cdns,usb3 > > reg: > + minItems: 2 > items: > - description: OTG controller registers > - description: XHCI Host controller registers > - description: DEVICE controller registers This is not accurate anymore, given you're allowing no otg registers for the no_drd case. Probably should have both items lists here, one for each configuration. > > reg-names: > + minItems: 2 > items: > - - const: otg > - - const: xhci > - - const: dev > + - enum: [ otg, xhci, dev ] > + - enum: [ otg, xhci, dev ] > + - enum: [ otg, xhci, dev ] If you delete the -, you can get away with only having one instance of the enum. You'd need to set maxItems: 3 though. > > interrupts: > - minItems: 3 > + minItems: 2 > items: > - description: XHCI host controller interrupt > - description: Device controller interrupt > @@ -41,12 +43,12 @@ properties: > cleared by xhci core, this interrupt is optional > > interrupt-names: > - minItems: 3 > + minItems: 2 > items: > - - const: host > - - const: peripheral > - - const: otg > - - const: wakeup > + - enum: [ host, peripheral, otg, wakeup ] > + - enum: [ host, peripheral, otg, wakeup ] > + - enum: [ host, peripheral, otg, wakeup ] > + - enum: [ host, peripheral, otg, wakeup ] I don't understand this. otg and wakeup are the ones you're making optional, so this list could remain as it was? > > port: > $ref: /schemas/graph.yaml#/properties/port > @@ -79,6 +81,13 @@ properties: > description: Enable resetting of PHY if Rx fail is detected > type: boolean > > + no_drd: On v1, I mistakenly thought this was a common property, but seems to not be. In that case, replace the _ with a - and add a vendor prefix. > + description: > + Indicates that the Dual-Role Device (DRD) register block is not > + implemented or is inaccessible. In this case, the controller > + must operate in a fixed peripheral or host mode. > + type: boolean > + > dependencies: > port: [ usb-role-switch ] > > @@ -93,6 +102,49 @@ allOf: > - $ref: usb-drd.yaml# > - $ref: usb-xhci.yaml# > > + - if: > + properties: > + no_drd: true Drop this, it's not needed. "required: - no_drd" does it for you. > + required: > + - no_drd > + then: > + required: > + - dr_mode And drop this too. pw-bot: changes-requested Thanks, Conor. > + properties: > + reg: > + maxItems: 2 > + reg-names: > + items: > + - const: xhci > + - const: dev > + interrupts: > + maxItems: 2 > + interrupt-names: > + items: > + - const: host > + - const: peripheral > + dr_mode: > + enum: [host, peripheral] > + else: > + properties: > + reg: > + minItems: 3 > + reg-names: > + items: > + - const: otg > + - const: xhci > + - const: dev > + interrupts: > + minItems: 3 > + maxItems: 4 > + interrupt-names: > + minItems: 3 > + items: > + - const: host > + - const: peripheral > + - const: otg > + - const: wakeup > + > unevaluatedProperties: false > > examples: > > -- > 2.43.0 > > [-- Attachment #2: signature.asc --] [-- Type: application/pgp-signature, Size: 228 bytes --] ^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH v6 1/2] dt-bindings: usb: cdns3: Add no_drd property 2026-05-13 11:44 ` [PATCH v6 1/2] dt-bindings: usb: cdns3: Add no_drd property Pawel Laszczak via B4 Relay 2026-05-13 18:59 ` Conor Dooley @ 2026-05-14 4:27 ` sashiko-bot 1 sibling, 0 replies; 6+ messages in thread From: sashiko-bot @ 2026-05-14 4:27 UTC (permalink / raw) To: Pawel Laszczak via B4 Relay; +Cc: krzk+dt, conor+dt, robh, devicetree Thank you for your contribution! Sashiko AI review found 5 potential issue(s) to consider: - [Medium] The schema enforces strict positional ordering for 'reg-names' and 'interrupt-names', directly contradicting the commit message's claim of 'flexible ordering'. - [Medium] The `reg` array positional descriptions in the base schema incorrectly document the XHCI registers as OTG registers when `no_drd` is true. - [Medium] The `else` block is missing a `minItems: 3` constraint for `reg-names`, allowing mismatched array lengths between `reg` and `reg-names`. - [Medium] The proposed property `no_drd` violates standard Device Tree naming conventions by using an underscore and missing a vendor prefix. - [Low] Introducing a redundant boolean property (`no_drd`) to indicate a missing hardware block is a hardware description anti-pattern. -- commit ef0683895d184cd4e0adacbed88f838da9544b55 Author: Pawel Laszczak <pawell@cadence.com> dt-bindings: usb: cdns3: Add no_drd property This commit introduces a new boolean property no_drd for Cadence USBSS/USBSSP controllers. It updates the schema to adjust reg, interrupts, and dr_mode requirements based on the presence of this property. > To achieve this, the schema is updated with an if-then-else logic > and 'reg-names'/'interrupt-names' use enums to allow flexible > ordering during validation. Does the if-then-else logic actually allow flexible ordering? Looking at the conditional blocks below, the items are defined as a list of const constraints. In JSON Schema, providing an array of schemas to items enforces strict positional validation, meaning the elements must appear in the exact order specified. > diff --git a/Documentation/devicetree/bindings/usb/cdns,usb3.yaml b/Documentation/devicetree/bindings/usb/cdns,usb3.yaml > index 2d95fb7321aff..717892a05dcdf 100644 > --- a/Documentation/devicetree/bindings/usb/cdns,usb3.yaml > +++ b/Documentation/devicetree/bindings/usb/cdns,usb3.yaml > @@ -20,19 +20,21 @@ properties: > const: cdns,usb3 > > reg: > + minItems: 2 > items: > - description: OTG controller registers > - description: XHCI Host controller registers > - description: DEVICE controller registers Does this cause an issue when the new property is set to true? When restricted to 2 items in the then block below, the first item must be xhci. However, the base schema defines the first item description as OTG controller registers. It seems the XHCI register at index 0 will be validated against and incorrectly described by the inherited OTG description. Fixed positional descriptions generally cannot be used if elements can be omitted from the beginning of a sequence. [ ... ] > @@ -79,6 +81,13 @@ properties: > description: Enable resetting of PHY if Rx fail is detected > type: boolean > > + no_drd: This isn't a bug, but device tree property names should generally use lowercase alphanumeric characters and hyphens instead of underscores. Would it also be better to include a vendor prefix to avoid namespace collisions, such as cdns,no-drd? > + description: > + Indicates that the Dual-Role Device (DRD) register block is not > + implemented or is inaccessible. In this case, the controller > + must operate in a fixed peripheral or host mode. > + type: boolean Is a dedicated boolean flag needed here? It seems the hardware configuration could be natively deduced by simply checking if otg is absent from reg-names. Relying on a negative feature flag instead of resource presence is usually discouraged in hardware descriptions. [ ... ] > + else: > + properties: > + reg: > + minItems: 3 > + reg-names: > + items: > + - const: otg > + - const: xhci > + - const: dev Could this allow mismatched array lengths between reg and reg-names? The reg property explicitly requires minItems: 3 here, but reg-names is not given a corresponding minItems constraint in this block. It will inherit minItems: 2 from the base schema, which might allow validation to succeed if a device tree defines 3 reg entries but only 2 reg-names. -- Sashiko AI review · https://sashiko.dev/#/patchset/20260513-b4-no_drd_config-v6-0-e7789cd0e581@cadence.com?part=1 ^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH v6 2/2] usb: cdnsp: Add support for device-only configuration 2026-05-13 11:44 [PATCH v6 0/2] usb: cdns3: support configurations without DRD block Pawel Laszczak via B4 Relay 2026-05-13 11:44 ` [PATCH v6 1/2] dt-bindings: usb: cdns3: Add no_drd property Pawel Laszczak via B4 Relay @ 2026-05-13 11:44 ` Pawel Laszczak via B4 Relay 2026-05-14 5:08 ` sashiko-bot 1 sibling, 1 reply; 6+ messages in thread From: Pawel Laszczak via B4 Relay @ 2026-05-13 11:44 UTC (permalink / raw) To: Peter Chen, Roger Quadros, Greg Kroah-Hartman, Rob Herring, Krzysztof Kozlowski, Conor Dooley Cc: linux-usb, devicetree, linux-kernel, Pawel Laszczak, Bjorn Helgaas From: Pawel Laszczak <pawell@cadence.com> This patch introduces support for the Cadence USBSSP (cdnsp) controller in hardware configurations where the Dual-Role Device (DRD) register block is not implemented or is inaccessible. In such cases, the driver cannot rely on the DRD logic to manage roles and must operate exclusively in a fixed peripheral/host mode. The change in BAR indexing (from BAR 2 to BAR 1) is a direct consequence of the 32-bit addressing used in this specific DRD-disabled hardware layout, compared to the 64-bit addressing used in DRD-enabled configurations. Tested on a PCI platform with a hardware configuration that lacks DRD support. Platform-side changes are included to support the PCI glue layer's property injection to handle this specific layout. Acked-by: Peter Chen <peter.chen@kernel.org> Acked-by: Bjorn Helgaas <bhelgaas@google.com> Signed-off-by: Pawel Laszczak <pawell@cadence.com> --- drivers/usb/cdns3/cdns3-plat.c | 26 +++++++++++++---------- drivers/usb/cdns3/cdnsp-pci.c | 47 ++++++++++++++++++++++++++++++++++-------- drivers/usb/cdns3/core.c | 3 ++- drivers/usb/cdns3/core.h | 4 ++++ drivers/usb/cdns3/drd.c | 44 +++++++++++++++++++++++++++++++++++++-- 5 files changed, 101 insertions(+), 23 deletions(-) diff --git a/drivers/usb/cdns3/cdns3-plat.c b/drivers/usb/cdns3/cdns3-plat.c index 3fe3109a3688..86c963a072db 100644 --- a/drivers/usb/cdns3/cdns3-plat.c +++ b/drivers/usb/cdns3/cdns3-plat.c @@ -81,6 +81,7 @@ static int cdns3_plat_probe(struct platform_device *pdev) if (cdns->pdata && cdns->pdata->override_apb_timeout) cdns->override_apb_timeout = cdns->pdata->override_apb_timeout; + cdns->no_drd = device_property_read_bool(dev, "no_drd"); platform_set_drvdata(pdev, cdns); ret = platform_get_irq_byname(pdev, "host"); @@ -113,21 +114,24 @@ static int cdns3_plat_probe(struct platform_device *pdev) cdns->dev_regs = regs; - cdns->otg_irq = platform_get_irq_byname(pdev, "otg"); - if (cdns->otg_irq < 0) - return dev_err_probe(dev, cdns->otg_irq, - "Failed to get otg IRQ\n"); - - res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "otg"); - if (!res) { - dev_err(dev, "couldn't get otg resource\n"); - return -ENXIO; + if (!cdns->no_drd) { + cdns->otg_irq = platform_get_irq_byname(pdev, "otg"); + if (cdns->otg_irq < 0) + return dev_err_probe(dev, cdns->otg_irq, + "Failed to get otg IRQ\n"); + + res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "otg"); + if (!res) { + dev_err(dev, "couldn't get otg resource\n"); + return -ENXIO; + } + cdns->otg_res = *res; + } else { + dev_dbg(dev, "No DRD support\n"); } cdns->phyrst_a_enable = device_property_read_bool(dev, "cdns,phyrst-a-enable"); - cdns->otg_res = *res; - cdns->wakeup_irq = platform_get_irq_byname_optional(pdev, "wakeup"); if (cdns->wakeup_irq == -EPROBE_DEFER) return cdns->wakeup_irq; diff --git a/drivers/usb/cdns3/cdnsp-pci.c b/drivers/usb/cdns3/cdnsp-pci.c index 432007cfe695..feb916229870 100644 --- a/drivers/usb/cdns3/cdnsp-pci.c +++ b/drivers/usb/cdns3/cdnsp-pci.c @@ -19,6 +19,7 @@ struct cdnsp_wrap { struct platform_device *plat_dev; + struct property_entry prop[3]; struct resource dev_res[6]; int devfn; }; @@ -29,10 +30,15 @@ struct cdnsp_wrap { #define RES_HOST_ID 3 #define RES_DEV_ID 4 #define RES_DRD_ID 5 - +/* DRD PCI configuration - 64-bit addressing */ +/* First PCI function */ #define PCI_BAR_HOST 0 -#define PCI_BAR_OTG 0 #define PCI_BAR_DEV 2 +/* Second PCI function */ +#define PCI_BAR_OTG 0 +/* Device only PCI configuration - 32-bit addressing */ +/* First PCI function */ +#define PCI_BAR_ONLY_DEV 1 #define PCI_DEV_FN_HOST_DEVICE 0 #define PCI_DEV_FN_OTG 1 @@ -40,6 +46,7 @@ struct cdnsp_wrap { #define PCI_DRIVER_NAME "cdns-pci-usbssp" #define PLAT_DRIVER_NAME "cdns-usb3" +#define PCI_DEVICE_ID_CDNS_UDC_USBSSP 0x0400 #define CHICKEN_APB_TIMEOUT_VALUE 0x1C20 static struct pci_dev *cdnsp_get_second_fun(struct pci_dev *pdev) @@ -65,6 +72,7 @@ static int cdnsp_pci_probe(struct pci_dev *pdev, struct cdnsp_wrap *wrap; struct resource *res; struct pci_dev *func; + bool no_drd = false; int ret = 0; /* @@ -75,11 +83,14 @@ static int cdnsp_pci_probe(struct pci_dev *pdev, pdev->devfn != PCI_DEV_FN_OTG)) return -EINVAL; + if (pdev->device == PCI_DEVICE_ID_CDNS_UDC_USBSSP) + no_drd = true; + func = cdnsp_get_second_fun(pdev); - if (!func) + if (!func && !no_drd) return -EINVAL; - if (func->class == PCI_CLASS_SERIAL_USB_XHCI || + if ((func && func->class == PCI_CLASS_SERIAL_USB_XHCI) || pdev->class == PCI_CLASS_SERIAL_USB_XHCI) { ret = -EINVAL; goto put_pci; @@ -93,7 +104,7 @@ static int cdnsp_pci_probe(struct pci_dev *pdev, pci_set_master(pdev); - if (pci_is_enabled(func)) { + if (func && pci_is_enabled(func)) { wrap = pci_get_drvdata(func); } else { wrap = kzalloc_obj(*wrap); @@ -106,10 +117,12 @@ static int cdnsp_pci_probe(struct pci_dev *pdev, res = wrap->dev_res; if (pdev->devfn == PCI_DEV_FN_HOST_DEVICE) { + int bar_dev = no_drd ? PCI_BAR_ONLY_DEV : PCI_BAR_DEV; + /* Function 0: host(BAR_0) + device(BAR_2). */ dev_dbg(&pdev->dev, "Initialize Device resources\n"); - res[RES_DEV_ID].start = pci_resource_start(pdev, PCI_BAR_DEV); - res[RES_DEV_ID].end = pci_resource_end(pdev, PCI_BAR_DEV); + res[RES_DEV_ID].start = pci_resource_start(pdev, bar_dev); + res[RES_DEV_ID].end = pci_resource_end(pdev, bar_dev); res[RES_DEV_ID].name = "dev"; res[RES_DEV_ID].flags = IORESOURCE_MEM; dev_dbg(&pdev->dev, "USBSSP-DEV physical base addr: %pa\n", @@ -145,9 +158,20 @@ static int cdnsp_pci_probe(struct pci_dev *pdev, wrap->dev_res[RES_IRQ_OTG_ID].flags = IORESOURCE_IRQ; } - if (pci_is_enabled(func)) { + if (no_drd || pci_is_enabled(func)) { + u8 idx = 0; + /* set up platform device info */ pdata.override_apb_timeout = CHICKEN_APB_TIMEOUT_VALUE; + + if (no_drd) { + wrap->prop[idx++] = PROPERTY_ENTRY_STRING("dr_mode", "peripheral"); + wrap->prop[idx++] = PROPERTY_ENTRY_BOOL("no_drd"); + } else { + wrap->prop[idx++] = PROPERTY_ENTRY_STRING("dr_mode", "otg"); + wrap->prop[idx++] = PROPERTY_ENTRY_BOOL("usb-role-switch"); + } + memset(&plat_info, 0, sizeof(plat_info)); plat_info.parent = &pdev->dev; plat_info.fwnode = pdev->dev.fwnode; @@ -158,6 +182,7 @@ static int cdnsp_pci_probe(struct pci_dev *pdev, plat_info.dma_mask = pdev->dma_mask; plat_info.data = &pdata; plat_info.size_data = sizeof(pdata); + plat_info.properties = wrap->prop; wrap->devfn = pdev->devfn; /* register platform device */ wrap->plat_dev = platform_device_register_full(&plat_info); @@ -185,13 +210,17 @@ static void cdnsp_pci_remove(struct pci_dev *pdev) if (wrap->devfn == pdev->devfn) platform_device_unregister(wrap->plat_dev); - if (!pci_is_enabled(func)) + if (!func || !pci_is_enabled(func)) kfree(wrap); pci_dev_put(func); } static const struct pci_device_id cdnsp_pci_ids[] = { + { PCI_DEVICE(PCI_VENDOR_ID_CDNS, PCI_DEVICE_ID_CDNS_UDC_USBSSP), + .class = PCI_CLASS_SERIAL_USB_DEVICE }, + { PCI_DEVICE(PCI_VENDOR_ID_CDNS, PCI_DEVICE_ID_CDNS_UDC_USBSSP), + .class = PCI_CLASS_SERIAL_USB_CDNS }, { PCI_DEVICE(PCI_VENDOR_ID_CDNS, PCI_DEVICE_ID_CDNS_USBSSP), .class = PCI_CLASS_SERIAL_USB_DEVICE }, { PCI_DEVICE(PCI_VENDOR_ID_CDNS, PCI_DEVICE_ID_CDNS_USBSSP), diff --git a/drivers/usb/cdns3/core.c b/drivers/usb/cdns3/core.c index 6a8d1fefbc0d..504bdf13ea80 100644 --- a/drivers/usb/cdns3/core.c +++ b/drivers/usb/cdns3/core.c @@ -70,7 +70,8 @@ static void cdns_role_stop(struct cdns *cdns) static void cdns_exit_roles(struct cdns *cdns) { cdns_role_stop(cdns); - cdns_drd_exit(cdns); + if (!cdns->no_drd) + cdns_drd_exit(cdns); } /** diff --git a/drivers/usb/cdns3/core.h b/drivers/usb/cdns3/core.h index bca973b999a4..8c492fda924c 100644 --- a/drivers/usb/cdns3/core.h +++ b/drivers/usb/cdns3/core.h @@ -84,6 +84,9 @@ struct cdns3_platform_data { * value in CHICKEN_BITS_3 will be preserved. * @gadget_init: pointer to gadget initialization function * @host_init: pointer to host initialization function + * @no_drd: DRD register block is inaccessible. The controller is hardwired to + * single role (host or device) or the logic for role switching is + * missing. */ struct cdns { struct device *dev; @@ -124,6 +127,7 @@ struct cdns { u32 override_apb_timeout; int (*gadget_init)(struct cdns *cdns); int (*host_init)(struct cdns *cdns); + bool no_drd; }; int cdns_hw_role_switch(struct cdns *cdns); diff --git a/drivers/usb/cdns3/drd.c b/drivers/usb/cdns3/drd.c index 84fb38a5723a..f87cf85cb97a 100644 --- a/drivers/usb/cdns3/drd.c +++ b/drivers/usb/cdns3/drd.c @@ -87,6 +87,9 @@ int cdns_get_id(struct cdns *cdns) { int id; + if (cdns->no_drd) + return 0; + id = readl(&cdns->otg_regs->sts) & OTGSTS_ID_VALUE; dev_dbg(cdns->dev, "OTG ID: %d", id); @@ -107,7 +110,7 @@ void cdns_clear_vbus(struct cdns *cdns) { u32 reg; - if (cdns->version != CDNSP_CONTROLLER_V2) + if (cdns->version != CDNSP_CONTROLLER_V2 || cdns->no_drd) return; reg = readl(&cdns->otg_cdnsp_regs->override); @@ -120,7 +123,7 @@ void cdns_set_vbus(struct cdns *cdns) { u32 reg; - if (cdns->version != CDNSP_CONTROLLER_V2) + if (cdns->version != CDNSP_CONTROLLER_V2 || cdns->no_drd) return; reg = readl(&cdns->otg_cdnsp_regs->override); @@ -181,6 +184,9 @@ int cdns_drd_host_on(struct cdns *cdns) u32 val, ready_bit; int ret; + if (cdns->no_drd) + goto phy_set; + /* Enable host mode. */ writel(OTGCMD_HOST_BUS_REQ | OTGCMD_OTG_DIS, &cdns->otg_regs->cmd); @@ -197,6 +203,7 @@ int cdns_drd_host_on(struct cdns *cdns) if (ret) dev_err(cdns->dev, "timeout waiting for xhci_ready\n"); +phy_set: phy_set_mode(cdns->usb2_phy, PHY_MODE_USB_HOST); phy_set_mode(cdns->usb3_phy, PHY_MODE_USB_HOST); return ret; @@ -210,6 +217,9 @@ void cdns_drd_host_off(struct cdns *cdns) { u32 val; + if (cdns->no_drd) + goto phy_set; + writel(OTGCMD_HOST_BUS_DROP | OTGCMD_DEV_BUS_DROP | OTGCMD_DEV_POWER_OFF | OTGCMD_HOST_POWER_OFF, &cdns->otg_regs->cmd); @@ -218,6 +228,8 @@ void cdns_drd_host_off(struct cdns *cdns) readl_poll_timeout_atomic(&cdns->otg_regs->state, val, !(val & OTGSTATE_HOST_STATE_MASK), 1, 2000000); + +phy_set: phy_set_mode(cdns->usb2_phy, PHY_MODE_INVALID); phy_set_mode(cdns->usb3_phy, PHY_MODE_INVALID); } @@ -234,6 +246,9 @@ int cdns_drd_gadget_on(struct cdns *cdns) u32 ready_bit; int ret, val; + if (cdns->no_drd) + goto phy_set; + /* switch OTG core */ writel(OTGCMD_DEV_BUS_REQ | reg, &cdns->otg_regs->cmd); @@ -251,6 +266,7 @@ int cdns_drd_gadget_on(struct cdns *cdns) return ret; } +phy_set: phy_set_mode(cdns->usb2_phy, PHY_MODE_USB_DEVICE); phy_set_mode(cdns->usb3_phy, PHY_MODE_USB_DEVICE); return 0; @@ -265,6 +281,9 @@ void cdns_drd_gadget_off(struct cdns *cdns) { u32 val; + if (cdns->no_drd) + goto phy_set; + /* * Driver should wait at least 10us after disabling Device * before turning-off Device (DEV_BUS_DROP). @@ -277,6 +296,8 @@ void cdns_drd_gadget_off(struct cdns *cdns) readl_poll_timeout_atomic(&cdns->otg_regs->state, val, !(val & OTGSTATE_DEV_STATE_MASK), 1, 2000000); + +phy_set: phy_set_mode(cdns->usb2_phy, PHY_MODE_INVALID); phy_set_mode(cdns->usb3_phy, PHY_MODE_INVALID); } @@ -392,6 +413,19 @@ int cdns_drd_init(struct cdns *cdns) u32 state, reg; int ret; + if (cdns->no_drd) { + cdns->dr_mode = usb_get_dr_mode(cdns->dev); + cdns->version = CDNSP_CONTROLLER_V2; + + if (cdns->dr_mode != USB_DR_MODE_HOST && + cdns->dr_mode != USB_DR_MODE_PERIPHERAL) { + dev_err(cdns->dev, "Incorrect dr_mode\n"); + return -EINVAL; + } + + return 0; + } + regs = devm_ioremap_resource(cdns->dev, &cdns->otg_res); if (IS_ERR(regs)) return PTR_ERR(regs); @@ -492,6 +526,9 @@ int cdns_drd_init(struct cdns *cdns) int cdns_drd_exit(struct cdns *cdns) { + if (cdns->no_drd) + return 0; + cdns_otg_disable_irq(cdns); return 0; @@ -500,6 +537,9 @@ int cdns_drd_exit(struct cdns *cdns) /* Indicate the cdns3 core was power lost before */ bool cdns_power_is_lost(struct cdns *cdns) { + if (cdns->no_drd) + return false; + if (cdns->version == CDNS3_CONTROLLER_V0) { if (!(readl(&cdns->otg_v0_regs->simulate) & BIT(0))) return true; -- 2.43.0 ^ permalink raw reply related [flat|nested] 6+ messages in thread
* Re: [PATCH v6 2/2] usb: cdnsp: Add support for device-only configuration 2026-05-13 11:44 ` [PATCH v6 2/2] usb: cdnsp: Add support for device-only configuration Pawel Laszczak via B4 Relay @ 2026-05-14 5:08 ` sashiko-bot 0 siblings, 0 replies; 6+ messages in thread From: sashiko-bot @ 2026-05-14 5:08 UTC (permalink / raw) To: Pawel Laszczak via B4 Relay; +Cc: krzk+dt, conor+dt, robh, devicetree Thank you for your contribution! Sashiko AI review found 3 potential issue(s) to consider: - [High] Uninitialized variable `ret` is returned from `cdns_drd_host_on` when `cdns->no_drd` is true. - [High] Error path in `cdnsp_pci_probe()` conditionally frees a shared `wrap` structure, leading to a use-after-free and double-free vulnerability. - [High] Concurrent execution of `cdnsp_pci_probe()` for both PCI functions causes a NULL pointer dereference due to lockless synchronization. -- commit ced8f360cae2b8973233c3e0a1eebbd2eb3a4cc6 Author: Pawel Laszczak <pawell@cadence.com> usb: cdnsp: Add support for device-only configuration This commit introduces support for the Cadence USBSSP (cdnsp) controller in hardware configurations where the Dual-Role Device (DRD) register block is not implemented or inaccessible, forcing the driver to rely on a fixed peripheral or host mode. > diff --git a/drivers/usb/cdns3/cdnsp-pci.c b/drivers/usb/cdns3/cdnsp-pci.c > index 432007cfe6954..feb9162298709 100644 > --- a/drivers/usb/cdns3/cdnsp-pci.c > +++ b/drivers/usb/cdns3/cdnsp-pci.c > @@ -93,7 +104,7 @@ static int cdnsp_pci_probe(struct pci_dev *pdev, > > pci_set_master(pdev); > > - if (pci_is_enabled(func)) { > + if (func && pci_is_enabled(func)) { > wrap = pci_get_drvdata(func); > } else { > wrap = kzalloc_obj(*wrap); Is it possible for the Host and OTG PCI functions to be probed concurrently? If the first thread observes pci_is_enabled(func) as true because the second thread just enabled it, but the second thread hasn't yet reached pci_set_drvdata() at the end of its probe, could pci_get_drvdata() return NULL here? This would cause a NULL pointer dereference when accessing wrap->dev_res further down. [ ... ] > @@ -158,6 +182,7 @@ static int cdnsp_pci_probe(struct pci_dev *pdev, > plat_info.dma_mask = pdev->dma_mask; > plat_info.data = &pdata; > plat_info.size_data = sizeof(pdata); > + plat_info.properties = wrap->prop; > wrap->devfn = pdev->devfn; > /* register platform device */ > wrap->plat_dev = platform_device_register_full(&plat_info); If the second function (OTG) retrieves a shared wrap structure allocated by the first function (Host), and platform_device_register_full() fails, the error path will execute kfree(wrap). Does this create a dangling pointer in the first function's drvdata, leading to a use-after-free or double-free later when the first function is removed? [ ... ] > diff --git a/drivers/usb/cdns3/drd.c b/drivers/usb/cdns3/drd.c > index 84fb38a5723a3..f87cf85cb97a1 100644 > --- a/drivers/usb/cdns3/drd.c > +++ b/drivers/usb/cdns3/drd.c > @@ -181,6 +184,9 @@ int cdns_drd_host_on(struct cdns *cdns) > u32 val, ready_bit; > int ret; > > + if (cdns->no_drd) > + goto phy_set; > + > /* Enable host mode. */ > writel(OTGCMD_HOST_BUS_REQ | OTGCMD_OTG_DIS, [ ... ] > @@ -197,6 +203,7 @@ int cdns_drd_host_on(struct cdns *cdns) > if (ret) > dev_err(cdns->dev, "timeout waiting for xhci_ready\n"); > > +phy_set: > phy_set_mode(cdns->usb2_phy, PHY_MODE_USB_HOST); > phy_set_mode(cdns->usb3_phy, PHY_MODE_USB_HOST); > return ret; When cdns->no_drd is true, execution jumps to the phy_set label, bypassing the initialization of the local variable ret. Could this return uninitialized stack memory and cause callers like cdns_role_start() to see a spurious failure and unexpectedly abort the host role startup? -- Sashiko AI review · https://sashiko.dev/#/patchset/20260513-b4-no_drd_config-v6-0-e7789cd0e581@cadence.com?part=2 ^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2026-05-14 5:08 UTC | newest] Thread overview: 6+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2026-05-13 11:44 [PATCH v6 0/2] usb: cdns3: support configurations without DRD block Pawel Laszczak via B4 Relay 2026-05-13 11:44 ` [PATCH v6 1/2] dt-bindings: usb: cdns3: Add no_drd property Pawel Laszczak via B4 Relay 2026-05-13 18:59 ` Conor Dooley 2026-05-14 4:27 ` sashiko-bot 2026-05-13 11:44 ` [PATCH v6 2/2] usb: cdnsp: Add support for device-only configuration Pawel Laszczak via B4 Relay 2026-05-14 5:08 ` sashiko-bot
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox