[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20170913202534.vf4r7dmuyfectbmv@rob-hp-laptop>
Date: Wed, 13 Sep 2017 15:25:34 -0500
From: Rob Herring <robh@...nel.org>
To: Ravi Shankar Jonnalagadda <venkata.ravi.jonnalagadda@...inx.com>
Cc: vinod.koul@...el.com, mark.rutland@....com,
michal.simek@...inx.com, soren.brinkmann@...inx.com,
dan.j.williams@...el.com, bhelgaas@...gle.com, vjonnal@...inx.com,
lorenzo.pieralisi@....com, bharat.kumar.gogada@...inx.com,
dmaengine@...r.kernel.org, devicetree@...r.kernel.org,
linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
linux-pci@...r.kernel.org, rgummal@...inx.com
Subject: Re: [PATCH v2 5/5] devicetree: zynqmp_ps_pcie: Devicetree binding
for Root DMA
On Fri, Sep 08, 2017 at 05:53:08PM +0530, Ravi Shankar Jonnalagadda wrote:
> Binding explaining devicetree usage for enabling Root DMA capability
>
> Signed-off-by: Ravi Shankar Jonnalagadda <vjonnal@...inx.com>
> Signed-off-by: RaviKiran Gummaluri <rgummal@...inx.com>
> ---
> .../devicetree/bindings/dma/xilinx/ps-pcie-dma.txt | 67 ++++++++++++++++++++++
> 1 file changed, 67 insertions(+)
> create mode 100644 Documentation/devicetree/bindings/dma/xilinx/ps-pcie-dma.txt
>
> diff --git a/Documentation/devicetree/bindings/dma/xilinx/ps-pcie-dma.txt b/Documentation/devicetree/bindings/dma/xilinx/ps-pcie-dma.txt
> new file mode 100644
> index 0000000..1522a49
> --- /dev/null
> +++ b/Documentation/devicetree/bindings/dma/xilinx/ps-pcie-dma.txt
> @@ -0,0 +1,67 @@
> +* Xilinx PS PCIe Root DMA
> +
> +Required properties:
> +- compatible: Should be "xlnx,ps_pcie_dma-1.00.a"
> +- reg: Register offset for Root DMA channels
> +- reg-names: Name for the register. Should be "xlnx,ps_pcie_regbase"
*-names for a single entry is pointless.
> +- interrupts: Interrupt pin for Root DMA
> +- interrupt-names: Name for the interrupt. Should be "xlnx,ps_pcie_rootdma_intr"
ditto
> +- interrupt-parent: Should be gic in case of zynqmp
> +- xlnx,rootdma: Indicates this platform device is root dma.
> + This is required as the same platform driver will be invoked by pcie end points too
platform device and driver are Linux terms.
> +- xlnx,dma_vendorid: 16 bit PCIe device vendor id.
> + This can be later used by dma client for matching while using dma_request_channel
> +- xlnx,dma_deviceid: 16 bit PCIe device id
> + This can be later used by dma client for matching while using dma_request_channel
This is the id's of the client? Sounds like you should use the DMA
binding.
> +- xlnx,numchannels: Indicates number of channels to be enabled for the device.
> + Valid values are from 1 to 4 for zynqmp
DMA binding has a similar property.
> +- xlnx,ps_pcie_channel : One for each channel to be enabled.
s/_/-/
> + This array contains channel specific properties.
> + Index 0: Direction of channel
> + Direction of channel can be either PCIe Memory to AXI memory i.e., Host to Card or
> + AXI Memory to PCIe memory i.e., Card to Host
> + PCIe to AXI Channel Direction is represented as 0x1
> + AXI to PCIe Channel Direction is represented as 0x0
> + Index 1: Number of Buffer Descriptors
> + This number describes number of buffer descriptors to be allocated for a channel
> + Index 2: Number of Queues
> + Each Channel has four DMA Buffer Descriptor Queues.
> + By default All four Queues will be managed by Root DMA driver.
> + User may choose to have only two queues either Source and it's Status Queue or
> + Destination and it's Status Queue to be handled by Driver.
> + The other two queues need to be handled by user logic which will not be part of this driver.
> + All Queues on Host is represented by 0x4
> + Two Queues on Host is represented by 0x2
> + Index 3: Coaelse Count
> + This number indicates the number of transfers after which interrupt needs to
> + be raised for the particular channel. The allowed range is from 0 to 255
> + Index 4: Coaelse Count Timer frequency
> + This property is used to control the frequency of poll timer. Poll timer is
> + created for a channel whenever coalesce count value (>= 1) is programmed for the particular
> + channel. This timer is helpful in draining out completed transactions even though interrupt is
> + not generated.
> +
> +Client Usage:
> + DMA clients can request for these channels using dma_request_channel API
> +
> +
> +Xilinx PS PCIe Root DMA node Example
> +++++++++++++++++++++++++++++++++++++
> +
> + pci_rootdma: rootdma@...f0000 {
dma-controller@...
> + compatible = "xlnx,ps_pcie_dma-1.00.a";
> + reg = <0x0 0xfd0f0000 0x0 0x1000>;
> + reg-names = "xlnx,ps_pcie_regbase";
> + interrupts = <0 117 4>;
> + interrupt-names = "xlnx,ps_pcie_rootdma_intr";
> + interrupt-parent = <&gic>;
> + xlnx,rootdma;
> + xlnx,dma_vendorid = /bits/ 16 <0x10EE>;
> + xlnx,dma_deviceid = /bits/ 16 <0xD021>;
> + xlnx,numchannels = <0x4>;
> + #size-cells = <0x5>;
> + xlnx,ps_pcie_channel0 = <0x1 0x7CF 0x4 0x0 0x3E8>;
> + xlnx,ps_pcie_channel1 = <0x0 0x7CF 0x4 0x0 0x3E8>;
> + xlnx,ps_pcie_channel2 = <0x1 0x7CF 0x4 0x0 0x3E8>;
> + xlnx,ps_pcie_channel3 = <0x0 0x7CF 0x4 0x0 0x3E8>;
> + };
> --
> 2.7.4
>
> --
> To unsubscribe from this list: send the line "unsubscribe devicetree" in
> the body of a message to majordomo@...r.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists