[PATCH 0/3] Add Broadcom PDC mailbox driver

[PATCH 0/3] Add Broadcom PDC mailbox driver

[Rob Rice]

From: Rob Rice <rrice@broadcom.com>

The Broadcom PDC mailbox driver is a ring manager for the Broadcom
Secure Processing Unit (SPU) kernel driver and other mailbox clients.
The PDC driver registers with the Linux mailbox framework a mailbox
controller, once for each PDC instance. Ring 0 for each PDC is registered as
a mailbox channel. The PDC driver uses interrupts to determine when ring
transfers to and from an offload engine are complete. The PDC driver uses
threaded IRQs so that response messages are handled outside of interrupt
context.

Rob Rice (3):
  dt-bindings: add bindings documentation for PDC driver.
  mailbox: Add Broadcom PDC mailbox driver
  arm64: dts: Add Broadcom Northstar2 device tree entries for PDC
    driver.

 .../bindings/mailbox/brcm,iproc-pdc-mbox.txt       |   23 +
 arch/arm64/boot/dts/broadcom/ns2.dtsi              |   36 +
 drivers/mailbox/Kconfig                            |    9 +
 drivers/mailbox/Makefile                           |    2 +
 drivers/mailbox/mailbox-pdc/Makefile               |    8 +
 drivers/mailbox/mailbox-pdc/pdc.c                  | 1181 ++++++++++++++++++++
 drivers/mailbox/mailbox-pdc/pdc.h                  |  303 +++++
 drivers/mailbox/mailbox-pdc/pdc_debug.c            |  101 ++
 drivers/mailbox/mailbox-pdc/pdc_debug.h            |   24 +
 include/linux/mailbox/brcm-message.h               |   56 +
 10 files changed, 1743 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/mailbox/brcm,iproc-pdc-mbox.txt
 create mode 100644 drivers/mailbox/mailbox-pdc/Makefile
 create mode 100644 drivers/mailbox/mailbox-pdc/pdc.c
 create mode 100644 drivers/mailbox/mailbox-pdc/pdc.h
 create mode 100644 drivers/mailbox/mailbox-pdc/pdc_debug.c
 create mode 100644 drivers/mailbox/mailbox-pdc/pdc_debug.h
 create mode 100644 include/linux/mailbox/brcm-message.h

-- 
2.1.0

[PATCH 1/3] dt-bindings: add bindings documentation for PDC driver.

[Rob Rice]

From: Rob Rice <rrice@broadcom.com>

Add the device tree binding documentation for the PDC hardware
in Broadcom iProc SoCs.

Signed-off-by: Rob Rice <rob.rice@broadcom.com>
Reviewed-by: Ray Jui <ray.jui@broadcom.com>
Reviewed-by: Anup Patel <anup.patel@broadcom.com>
Reviewed-by: Scott Branden <scott.branden@broadcom.com>
---
 .../bindings/mailbox/brcm,iproc-pdc-mbox.txt       | 23 ++++++++++++++++++++++
 1 file changed, 23 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/mailbox/brcm,iproc-pdc-mbox.txt

diff --git a/Documentation/devicetree/bindings/mailbox/brcm,iproc-pdc-mbox.txt b/Documentation/devicetree/bindings/mailbox/brcm,iproc-pdc-mbox.txt
new file mode 100644
index 0000000..411ccf4
--- /dev/null
+++ b/Documentation/devicetree/bindings/mailbox/brcm,iproc-pdc-mbox.txt
@@ -0,0 +1,23 @@
+The PDC driver manages data transfer to and from various offload engines
+on some Broadcom SoCs. An SoC may have multiple PDC hardware blocks. There is
+one device tree entry per block.
+
+Required properties:
+- compatible : Should be "brcm,iproc-pdc-mbox".
+- reg: Should contain PDC registers location and length.
+- interrupts: Should contain the IRQ line for the PDC.
+- #mbox-cells: 1
+- brcm,rx-status-len: Length of metadata preceding received frames, in bytes.
+
+Optional properties:
+- brcm,use-bcm-hdr:  present if a BCM header precedes each frame.
+
+Example:
+	pdc0: iproc-pdc0@0x612c0000 {
+		compatible = "brcm,iproc-pdc-mbox";
+		reg = <0 0x612c0000 0 0x445>;  /* PDC FS0 regs */
+		interrupts = <GIC_SPI 187 IRQ_TYPE_LEVEL_HIGH>;
+		#mbox-cells = <1>;   /* one cell per mailbox channel */
+		brcm,rx-status-len = <32>;
+		brcm,use-bcm-hdr;
+	};
-- 
2.1.0

Re: [PATCH 1/3] dt-bindings: add bindings documentation for PDC driver.

[Rob Herring]

On Tue, May 24, 2016 at 02:07:28PM -0400, Rob Rice wrote:
> From: Rob Rice <rrice@broadcom.com>
> 
> Add the device tree binding documentation for the PDC hardware
> in Broadcom iProc SoCs.
> 
> Signed-off-by: Rob Rice <rob.rice@broadcom.com>
> Reviewed-by: Ray Jui <ray.jui@broadcom.com>
> Reviewed-by: Anup Patel <anup.patel@broadcom.com>
> Reviewed-by: Scott Branden <scott.branden@broadcom.com>
> ---
>  .../bindings/mailbox/brcm,iproc-pdc-mbox.txt       | 23 ++++++++++++++++++++++
>  1 file changed, 23 insertions(+)
>  create mode 100644 Documentation/devicetree/bindings/mailbox/brcm,iproc-pdc-mbox.txt

Acked-by: Rob Herring <robh@kernel.org>

[PATCH 2/3] mailbox: Add Broadcom PDC mailbox driver

[Rob Rice]

From: Rob Rice <rrice@broadcom.com>

The Broadcom PDC mailbox driver is a mailbox controller that
manages data transfers to and from one or more offload engines.

Signed-off-by: Rob Rice <rob.rice@broadcom.com>
Reviewed-by: Scott Branden <scott.branden@broadcom.com>
Reviewed-by: Ray Jui <ray.jui@broadcom.com>
---
 drivers/mailbox/Kconfig                 |    9 +
 drivers/mailbox/Makefile                |    2 +
 drivers/mailbox/mailbox-pdc/Makefile    |    8 +
 drivers/mailbox/mailbox-pdc/pdc.c       | 1181 +++++++++++++++++++++++++++++++
 drivers/mailbox/mailbox-pdc/pdc.h       |  303 ++++++++
 drivers/mailbox/mailbox-pdc/pdc_debug.c |  101 +++
 drivers/mailbox/mailbox-pdc/pdc_debug.h |   24 +
 include/linux/mailbox/brcm-message.h    |   56 ++
 8 files changed, 1684 insertions(+)
 create mode 100644 drivers/mailbox/mailbox-pdc/Makefile
 create mode 100644 drivers/mailbox/mailbox-pdc/pdc.c
 create mode 100644 drivers/mailbox/mailbox-pdc/pdc.h
 create mode 100644 drivers/mailbox/mailbox-pdc/pdc_debug.c
 create mode 100644 drivers/mailbox/mailbox-pdc/pdc_debug.h
 create mode 100644 include/linux/mailbox/brcm-message.h

diff --git a/drivers/mailbox/Kconfig b/drivers/mailbox/Kconfig
index 5305923..19c8c9a 100644
--- a/drivers/mailbox/Kconfig
+++ b/drivers/mailbox/Kconfig
@@ -123,4 +123,13 @@ config XGENE_SLIMPRO_MBOX
 	  It is used to send short messages between ARM64-bit cores and
 	  the SLIMpro Management Engine, primarily for PM. Say Y here if you
 	  want to use the APM X-Gene SLIMpro IPCM support.
+
+config BCM_PDC_MBOX
+	tristate "Broadcom PDC Mailbox"
+	depends on ARM64
+	default ARCH_BCM_IPROC
+	help
+	  Mailbox implementation for the Broadcom PDC ring manager,
+	  which provides access to various offload engines on Broadcom
+	  SoCs. Say Y here if you want to use the Broadcom PDC.
 endif
diff --git a/drivers/mailbox/Makefile b/drivers/mailbox/Makefile
index 0be3e74..2c14a03 100644
--- a/drivers/mailbox/Makefile
+++ b/drivers/mailbox/Makefile
@@ -25,3 +25,5 @@ obj-$(CONFIG_TI_MESSAGE_MANAGER) += ti-msgmgr.o
 obj-$(CONFIG_XGENE_SLIMPRO_MBOX) += mailbox-xgene-slimpro.o
 
 obj-$(CONFIG_HI6220_MBOX)	+= hi6220-mailbox.o
+
+obj-$(CONFIG_BCM_PDC_MBOX)	+= mailbox-pdc/
diff --git a/drivers/mailbox/mailbox-pdc/Makefile b/drivers/mailbox/mailbox-pdc/Makefile
new file mode 100644
index 0000000..5dc78c3
--- /dev/null
+++ b/drivers/mailbox/mailbox-pdc/Makefile
@@ -0,0 +1,8 @@
+# Makefile for Broadcom PDC Mailbox driver
+
+# Uncomment to enable debug tracing in the PDC driver
+# CFLAGS_pdc.o := -DDEBUG
+# CFLAGS_pdc_debug.o := -DDEBUG
+
+pdc-mbox-objs := pdc.o pdc_debug.o
+obj-$(CONFIG_BCM_PDC_MBOX) := pdc-mbox.o
diff --git a/drivers/mailbox/mailbox-pdc/pdc.c b/drivers/mailbox/mailbox-pdc/pdc.c
new file mode 100644
index 0000000..1c24ea8
--- /dev/null
+++ b/drivers/mailbox/mailbox-pdc/pdc.c
@@ -0,0 +1,1181 @@
+/*
+ * Copyright 2016 Broadcom
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, version 2, as
+ * published by the Free Software Foundation (the "GPL").
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License version 2 (GPLv2) for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * version 2 (GPLv2) along with this source code.
+ */
+
+/*
+ * Broadcom PDC Mailbox Driver
+ * The PDC provides a ring based programming interface to one or more hardware
+ * offload engines. For example, the PDC driver works with both SPU-M and SPU2
+ * cryptographic offload hardware. In some chips the PDC is referred to as MDE.
+ *
+ * The PDC driver registers with the Linux mailbox framework as a mailbox
+ * controller, once for each PDC instance. Ring 0 for each PDC is registered as
+ * a mailbox channel. The PDC driver uses interrupts to determine when data
+ * transfers to and from an offload engine are complete. The PDC driver uses
+ * threaded IRQs so that response messages are handled outside of interrupt
+ * context.
+ *
+ * The PDC driver allows multiple messages to be pending in the descriptor
+ * rings. The tx_msg_start descriptor index indicates where the last message
+ * starts. The txin_numd value at this index indicates how many descriptor
+ * indexes make up the message. Similar state is kept on the receive side. When
+ * an rx interrupt indicates a response is ready, the PDC driver processes numd
+ * descriptors from the tx and rx ring, thus processing one response at a time.
+ */
+
+#include <linux/errno.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/interrupt.h>
+#include <linux/wait.h>
+#include <linux/platform_device.h>
+#include <linux/io.h>
+#include <linux/of.h>
+#include <linux/of_device.h>
+#include <linux/of_address.h>
+#include <linux/of_irq.h>
+#include <linux/mailbox_controller.h>
+#include <linux/mailbox/brcm-message.h>
+#include <linux/scatterlist.h>
+#include <linux/dma-direction.h>
+#include <linux/dma-mapping.h>
+#include <linux/dmapool.h>
+
+#include "pdc.h"
+#include "pdc_debug.h"
+
+/* Length of BCM header at start of SPU msg, in bytes */
+#define BCM_HDR_LEN  8
+
+/*
+ * PDC driver reserves ringset 0 on each SPU for its own use. The driver does
+ * not currently support use of multiple ringsets on a single PDC engine.
+ */
+#define PDC_RINGSET  0
+
+/*
+ * Interrupt mask and status definitions. Enable interrupts for tx and rx on
+ * ring 0
+ */
+#define PDC_XMTINT_0         (24 + PDC_RINGSET)
+#define PDC_RCVINT_0         (16 + PDC_RINGSET)
+#define PDC_XMTINTEN_0       BIT(PDC_XMTINT_0)
+#define PDC_RCVINTEN_0       BIT(PDC_RCVINT_0)
+#define PDC_INTMASK  (PDC_XMTINTEN_0 | PDC_RCVINTEN_0)
+#define PDC_LAZY_FRAMECOUNT  1
+#define PDC_LAZY_TIMEOUT     10000
+#define PDC_LAZY_INT  (PDC_LAZY_TIMEOUT | (PDC_LAZY_FRAMECOUNT << 24))
+#define PDC_INTMASK_OFFSET   0x24
+#define PDC_INTSTATUS_OFFSET 0x20
+#define PDC_RCVLAZY0_OFFSET  (0x30 + 4 * PDC_RINGSET)
+
+/*
+ * For SPU2, configure MDE_CKSUM_CONTROL to write 17 bytes of metadata
+ * before frame
+ */
+#define PDC_SPU2_RESP_HDR_LEN  17
+#define PDC_CKSUM_CTRL         BIT(27)
+#define PDC_CKSUM_CTRL_OFFSET  0x400
+
+#define PDC_SPUM_RESP_HDR_LEN  32
+
+/*
+ * Sets the following bits for write to transmit control reg:
+ *  0    - XmtEn - enable activity on the tx channel
+ * 11    - PtyChkDisable - parity check is disabled
+ * 20:18 - BurstLen = 3 -> 2^7 = 128 byte data reads from memory
+ */
+#define PDC_TX_CTL              0x000C0801
+
+/*
+ * Sets the following bits for write to receive control reg:
+ * 0     - RcvEn - enable activity on the rx channel
+ * 7:1   - RcvOffset - size in bytes of status region at start of rx frame buf
+ * 9     - SepRxHdrDescEn - place start of new frames only in descriptors
+ *                          that have StartOfFrame set
+ * 10    - OflowContinue - on rx FIFO overflow, clear rx fifo, discard all
+ *                         remaining bytes in current frame, report error
+ *                         in rx frame status for current frame
+ * 11    - PtyChkDisable - parity check is disabled
+ * 20:18 - BurstLen = 3 -> 2^7 = 128 byte data reads from memory
+ */
+#define PDC_RX_CTL              0x000C0E01
+
+#define CRYPTO_D64_RS0_CD_MASK   ((PDC_RING_ENTRIES * RING_ENTRY_SIZE) - 1)
+
+/* descriptor flags */
+#define D64_CTRL1_EOT   BIT(28)	/* end of descriptor table */
+#define D64_CTRL1_IOC   BIT(29)	/* interrupt on complete */
+#define D64_CTRL1_EOF   BIT(30)	/* end of frame */
+#define D64_CTRL1_SOF   BIT(31)	/* start of frame */
+
+#define RX_STATUS_OVERFLOW       0x00800000
+#define RX_STATUS_LEN            0x0000FFFF
+
+#define PDC_TXREGS_OFFSET  0x200
+#define PDC_RXREGS_OFFSET  0x220
+
+/* Maximum size buffer the DMA engine can handle */
+#define PDC_DMA_BUF_MAX 16384
+
+/* Global variables */
+
+struct pdc_globals {
+	/* Actual number of SPUs in hardware, as reported by device tree */
+	u32 num_spu;
+};
+
+static struct pdc_globals pdcg;
+
+/**
+ * pdc_build_rxd() - Build DMA descriptor to receive SPU result.
+ * @pdcs:      PDC state for SPU that will generate result
+ * @dma_addr:  DMA address of buffer that descriptor is being built for
+ * @buf_len:   Length of the receive buffer, in bytes
+ * @flags:     Flags to be stored in descriptor
+ */
+static inline void
+pdc_build_rxd(struct pdc_state *pdcs, dma_addr_t dma_addr,
+	      u32 buf_len, u32 flags)
+{
+	struct device *dev = &pdcs->pdev->dev;
+
+	dev_dbg(dev,
+		"Writing rx descriptor for PDC %u at index %u with length %u. flags %#x\n",
+		pdcs->pdc_idx, pdcs->rxout, buf_len, flags);
+
+	iowrite32(lower_32_bits(dma_addr),
+		  (void *)&pdcs->rxd_64[pdcs->rxout].addrlow);
+	iowrite32(upper_32_bits(dma_addr),
+		  (void *)&pdcs->rxd_64[pdcs->rxout].addrhigh);
+	iowrite32(flags, (void *)&pdcs->rxd_64[pdcs->rxout].ctrl1);
+	iowrite32(buf_len, (void *)&pdcs->rxd_64[pdcs->rxout].ctrl2);
+	/* bump ring index and return */
+	pdcs->rxout = NEXTRXD(pdcs->rxout, pdcs->nrxpost);
+}
+
+/**
+ * pdc_build_txd() - Build a DMA descriptor to transmit a SPU request to
+ * hardware.
+ * @pdcs:        PDC state for the SPU that will process this request
+ * @dma_addr:    DMA address of packet to be transmitted
+ * @buf_len:     Length of tx buffer, in bytes
+ * @flags:       Flags to be stored in descriptor
+ */
+static inline void
+pdc_build_txd(struct pdc_state *pdcs, dma_addr_t dma_addr, u32 buf_len,
+	      u32 flags)
+{
+	struct device *dev = &pdcs->pdev->dev;
+
+	dev_dbg(dev,
+		"Writing tx descriptor for PDC %u at index %u with length %u, flags %#x\n",
+		pdcs->pdc_idx, pdcs->txout, buf_len, flags);
+
+	iowrite32(lower_32_bits(dma_addr),
+		  (void *)&pdcs->txd_64[pdcs->txout].addrlow);
+	iowrite32(upper_32_bits(dma_addr),
+		  (void *)&pdcs->txd_64[pdcs->txout].addrhigh);
+	iowrite32(flags, (void *)&pdcs->txd_64[pdcs->txout].ctrl1);
+	iowrite32(buf_len, (void *)&pdcs->txd_64[pdcs->txout].ctrl2);
+
+	/* bump ring index and return */
+	pdcs->txout = NEXTTXD(pdcs->txout, pdcs->ntxpost);
+}
+
+/**
+ * pdc_receive() - Receive a response message from a given SPU.
+ * @pdcs:    PDC state for the SPU to receive from
+ * @mssg:    mailbox message to be returned to client
+ *
+ * When the return code indicates success, the response message is available in
+ * the receive buffers provided prior to submission of the request.
+ *
+ * Input:
+ *   pdcs - PDC state structure for the SPU to be polled
+ *   mssg - mailbox message to be returned to client. This function sets the
+ *	    context pointer on the message to help the client associate the
+ *	    response with a request.
+ *
+ * Return:  PDC_SUCCESS if one or more receive descriptors was processed
+ *          -EAGAIN indicates that no response message is available
+ *          -EIO an error occurred
+ */
+static int
+pdc_receive(struct pdc_state *pdcs, struct brcm_message *mssg)
+{
+	struct device *dev = &pdcs->pdev->dev;
+	u32 len, rx_status;
+	u32 num_frags;
+	int i;
+	u8 *resp_hdr;    /* virtual addr of start of resp message DMA header */
+	u32 frags_rdy;   /* number of fragments ready to read */
+	u32 rx_idx;      /* ring index of start of receive frame */
+	dma_addr_t resp_hdr_daddr;
+
+	spin_lock(&pdcs->pdc_lock);
+
+	/*
+	 * return if a complete response message is not yet ready.
+	 * rxin_numd[rxin] is the number of fragments in the next msg
+	 * to read.
+	 */
+	frags_rdy = NRXDACTIVE(pdcs->rxin, pdcs->last_rx_curr, pdcs->nrxpost);
+	if ((frags_rdy == 0) || (frags_rdy < pdcs->rxin_numd[pdcs->rxin])) {
+		/* See if the hw has written more fragments than we know */
+		pdcs->last_rx_curr =
+		    (ioread32((void *)&pdcs->rxregs_64->status0) &
+		     CRYPTO_D64_RS0_CD_MASK) / RING_ENTRY_SIZE;
+		frags_rdy = NRXDACTIVE(pdcs->rxin, pdcs->last_rx_curr,
+				       pdcs->nrxpost);
+		if ((frags_rdy == 0) ||
+		    (frags_rdy < pdcs->rxin_numd[pdcs->rxin])) {
+			/* No response ready */
+			spin_unlock(&pdcs->pdc_lock);
+			return -EAGAIN;
+		}
+		/* can't read descriptors/data until write index is read */
+		rmb();
+	}
+
+	num_frags = pdcs->txin_numd[pdcs->txin];
+	dma_unmap_sg(dev, pdcs->src_sg[pdcs->txin],
+		     sg_nents(pdcs->src_sg[pdcs->txin]), DMA_TO_DEVICE);
+
+	for (i = 0; i < num_frags; i++)
+		pdcs->txin = NEXTTXD(pdcs->txin, pdcs->ntxpost);
+
+	dev_dbg(dev, "PDC %u reclaimed %d tx descriptors",
+		pdcs->pdc_idx, num_frags);
+
+	rx_idx = pdcs->rxin;
+	num_frags = pdcs->rxin_numd[rx_idx];
+	/* Return opaque context with result */
+	mssg->ctx = pdcs->rxp_ctx[rx_idx];
+	pdcs->rxp_ctx[rx_idx] = NULL;
+	resp_hdr = pdcs->resp_hdr[rx_idx];
+	resp_hdr_daddr = pdcs->resp_hdr_daddr[rx_idx];
+	dma_unmap_sg(dev, pdcs->dst_sg[rx_idx],
+		     sg_nents(pdcs->dst_sg[rx_idx]), DMA_FROM_DEVICE);
+
+	for (i = 0; i < num_frags; i++)
+		pdcs->rxin = NEXTRXD(pdcs->rxin, pdcs->nrxpost);
+
+	spin_unlock(&pdcs->pdc_lock);
+
+	dev_dbg(dev, "PDC %u reclaimed %d rx descriptors",
+		pdcs->pdc_idx, num_frags);
+
+	dev_dbg(dev,
+		"PDC %u txin %u, txout %u, rxin %u, rxout %u, last_rx_curr %u\n",
+		pdcs->pdc_idx, pdcs->txin, pdcs->txout, pdcs->rxin,
+		pdcs->rxout, pdcs->last_rx_curr);
+
+	if (pdcs->pdc_resp_hdr_len == PDC_SPUM_RESP_HDR_LEN) {
+		/*
+		 * For SPU-M, get length of response msg and rx overflow status.
+		 */
+		rx_status = *((u32 *)resp_hdr);
+		len = rx_status & RX_STATUS_LEN;
+		dev_dbg(dev,
+			"SPU response length %u bytes", len);
+		if (unlikely(((rx_status & RX_STATUS_OVERFLOW) || (!len)))) {
+			if (rx_status & RX_STATUS_OVERFLOW) {
+				dev_err_ratelimited(dev,
+						    "crypto receive overflow");
+				pdcs->rx_oflow++;
+			} else {
+				dev_info_ratelimited(dev, "crypto rx len = 0");
+			}
+			return -EIO;
+		}
+	}
+
+	dma_pool_free(pdcs->rx_buf_pool, resp_hdr, resp_hdr_daddr);
+
+	pdcs->pdc_replies++;
+	/* if we read one or more rx descriptors, claim success */
+	if (num_frags > 0)
+		return PDC_SUCCESS;
+	else
+		return -EIO;
+}
+
+/**
+ * pdc_tx_list_sg_add() - Add the buffers in a scatterlist to the transmit
+ * descriptors for a given SPU. The scatterlist buffers contain the data for a
+ * SPU request message.
+ * @spu_idx:   The index of the SPU to submit the request to, [0, max_spu)
+ * @sg:        Scatterlist whose buffers contain part of the SPU request
+ *
+ * If a scatterlist buffer is larger than PDC_DMA_BUF_MAX, multiple descriptors
+ * are written for that buffer, each <= PDC_DMA_BUF_MAX byte in length.
+ *
+ * Return: PDC_SUCCESS if successful
+ *         < 0 otherwise
+ */
+static int pdc_tx_list_sg_add(struct pdc_state *pdcs, struct scatterlist *sg)
+{
+	u32 flags = 0;
+	u32 eot;
+	u32 tx_avail;
+
+	/*
+	 * Num descriptors needed. Conservatively assume we need a descriptor
+	 * for every entry in sg.
+	 */
+	u32 num_desc;
+	u32 desc_w = 0;	/* Number of tx descriptors written */
+	u32 bufcnt;	/* Number of bytes of buffer pointed to by descriptor */
+	dma_addr_t databufptr;	/* DMA address to put in descriptor */
+
+	num_desc = (u32)sg_nents(sg);
+
+	/* check whether enough tx descriptors are available */
+	tx_avail = pdcs->ntxpost - NTXDACTIVE(pdcs->txin, pdcs->txout,
+					      pdcs->ntxpost);
+	if (unlikely(num_desc > tx_avail)) {
+		pdcs->txnobuf++;
+		return -ENOSPC;
+	}
+
+	/* build tx descriptors */
+	if (pdcs->tx_msg_start == pdcs->txout) {
+		/* Start of frame */
+		pdcs->txin_numd[pdcs->tx_msg_start] = 0;
+		pdcs->src_sg[pdcs->txout] = sg;
+		flags = D64_CTRL1_SOF;
+	}
+
+	while (sg) {
+		if (unlikely(pdcs->txout == (pdcs->ntxd - 1)))
+			eot = D64_CTRL1_EOT;
+		else
+			eot = 0;
+
+		/*
+		 * If sg buffer larger than PDC limit, split across
+		 * multiple descriptors
+		 */
+		bufcnt = sg_dma_len(sg);
+		databufptr = sg_dma_address(sg);
+		while (bufcnt > PDC_DMA_BUF_MAX) {
+			pdc_build_txd(pdcs, databufptr, PDC_DMA_BUF_MAX,
+				      flags | eot);
+			desc_w++;
+			bufcnt -= PDC_DMA_BUF_MAX;
+			databufptr += PDC_DMA_BUF_MAX;
+			if (unlikely(pdcs->txout == (pdcs->ntxd - 1)))
+				eot = D64_CTRL1_EOT;
+			else
+				eot = 0;
+		}
+		sg = sg_next(sg);
+		if (!sg)
+			/* Writing last descriptor for frame */
+			flags |= (D64_CTRL1_EOF | D64_CTRL1_IOC);
+		pdc_build_txd(pdcs, databufptr, bufcnt, flags | eot);
+		desc_w++;
+		/* Clear start of frame after first descriptor */
+		flags &= ~D64_CTRL1_SOF;
+	}
+	pdcs->txin_numd[pdcs->tx_msg_start] += desc_w;
+
+	return PDC_SUCCESS;
+}
+
+/**
+ * pdc_tx_list_final() - Initiate DMA transfer of last frame written to tx
+ * ring.
+ * @pdcs:  PDC state for SPU to process the request
+ *
+ * Sets the index of the last descriptor written in both the rx and tx ring.
+ *
+ * Return: PDC_SUCCESS
+ */
+static int pdc_tx_list_final(struct pdc_state *pdcs)
+{
+	/*
+	 * write barrier to ensure all register writes are complete
+	 * before chip starts to process new request
+	 */
+	wmb();
+	iowrite32(pdcs->rxout << 4, (void *)&pdcs->rxregs_64->ptr);
+	iowrite32(pdcs->txout << 4, (void *)&pdcs->txregs_64->ptr);
+	pdcs->pdc_requests++;
+
+	return PDC_SUCCESS;
+}
+
+/**
+ * pdc_rx_list_init() - Start a new receive descriptor list for a given PDC.
+ * @pdcs:   PDC state for SPU handling request
+ * @dst_sg: scatterlist providing rx buffers for response to be returned to
+ *	    mailbox client
+ * @ctx:    Opaque context for this request
+ *
+ * Posts a single receive descriptor to hold the metadata that precedes a
+ * response. For example, with SPU-M, the metadata is a 32-byte DMA header and
+ * an 8-byte BCM header. Moves the msg_start descriptor indexes for both tx and
+ * rx to indicate the start of a new message.
+ *
+ * Return:  PDC_SUCCESS if successful
+ *          < 0 if an error (e.g., rx ring is full)
+ */
+static int pdc_rx_list_init(struct pdc_state *pdcs, struct scatterlist *dst_sg,
+			    void *ctx)
+{
+	u32 flags = 0;
+	u32 rx_avail;
+	u32 rx_pkt_cnt = 1;	/* Adding a single rx buffer */
+	dma_addr_t daddr;
+	void *vaddr;
+
+	rx_avail = pdcs->nrxpost - NRXDACTIVE(pdcs->rxin, pdcs->rxout,
+					      pdcs->nrxpost);
+	if (unlikely(rx_pkt_cnt > rx_avail)) {
+		pdcs->rxnobuf++;
+		return -ENOSPC;
+	}
+
+	/* allocate a buffer for the dma rx status */
+	vaddr = dma_pool_zalloc(pdcs->rx_buf_pool, GFP_ATOMIC, &daddr);
+	if (!vaddr)
+		return -ENOMEM;
+
+	/*
+	 * Update msg_start indexes for both tx and rx to indicate the start
+	 * of a new sequence of descriptor indexes that contain the fragments
+	 * of the same message.
+	 */
+	pdcs->rx_msg_start = pdcs->rxout;
+	pdcs->tx_msg_start = pdcs->txout;
+
+	/* This is always the first descriptor in the receive sequence */
+	flags = D64_CTRL1_SOF;
+	pdcs->rxin_numd[pdcs->rx_msg_start] = 1;
+
+	if (unlikely(pdcs->rxout == (pdcs->nrxd - 1)))
+		flags |= D64_CTRL1_EOT;
+
+	pdcs->rxp_ctx[pdcs->rxout] = ctx;
+	pdcs->dst_sg[pdcs->rxout] = dst_sg;
+	pdcs->resp_hdr[pdcs->rxout] = vaddr;
+	pdcs->resp_hdr_daddr[pdcs->rxout] = daddr;
+	pdc_build_rxd(pdcs, daddr, pdcs->pdc_resp_hdr_len, flags);
+	return PDC_SUCCESS;
+}
+
+/**
+ * pdc_rx_list_sg_add() - Add the buffers in a scatterlist to the receive
+ * descriptors for a given SPU. The caller must have already DMA mapped the
+ * scatterlist.
+ * @spu_idx:    Indicates which SPU the buffers are for
+ * @sg:         Scatterlist whose buffers are added to the receive ring
+ *
+ * If a receive buffer in the scatterlist is larger than PDC_DMA_BUF_MAX,
+ * multiple receive descriptors are written, each with a buffer <=
+ * PDC_DMA_BUF_MAX.
+ *
+ * Return: PDC_SUCCESS if successful
+ *         < 0 otherwise (e.g., receive ring is full)
+ */
+static int pdc_rx_list_sg_add(struct pdc_state *pdcs, struct scatterlist *sg)
+{
+	u32 flags = 0;
+	u32 rx_avail;
+
+	/*
+	 * Num descriptors needed. Conservatively assume we need a descriptor
+	 * for every entry from our starting point in the scatterlist.
+	 */
+	u32 num_desc;
+	u32 desc_w = 0;	/* Number of tx descriptors written */
+	u32 bufcnt;	/* Number of bytes of buffer pointed to by descriptor */
+	dma_addr_t databufptr;	/* DMA address to put in descriptor */
+
+	num_desc = (u32)sg_nents(sg);
+
+	rx_avail = pdcs->nrxpost - NRXDACTIVE(pdcs->rxin, pdcs->rxout,
+					      pdcs->nrxpost);
+	if (unlikely(num_desc > rx_avail)) {
+		pdcs->rxnobuf++;
+		return -ENOSPC;
+	}
+
+	while (sg) {
+		if (unlikely(pdcs->rxout == (pdcs->nrxd - 1)))
+			flags = D64_CTRL1_EOT;
+		else
+			flags = 0;
+
+		/*
+		 * If sg buffer larger than PDC limit, split across
+		 * multiple descriptors
+		 */
+		bufcnt = sg_dma_len(sg);
+		databufptr = sg_dma_address(sg);
+		while (bufcnt > PDC_DMA_BUF_MAX) {
+			pdc_build_rxd(pdcs, databufptr, PDC_DMA_BUF_MAX, flags);
+			desc_w++;
+			bufcnt -= PDC_DMA_BUF_MAX;
+			databufptr += PDC_DMA_BUF_MAX;
+			if (unlikely(pdcs->rxout == (pdcs->nrxd - 1)))
+				flags = D64_CTRL1_EOT;
+			else
+				flags = 0;
+		}
+		pdc_build_rxd(pdcs, databufptr, bufcnt, flags);
+		desc_w++;
+		sg = sg_next(sg);
+	}
+	pdcs->rxin_numd[pdcs->rx_msg_start] += desc_w;
+
+	return PDC_SUCCESS;
+}
+
+/**
+ * pdc_irq_handler() - Interrupt handler called in interrupt context.
+ * @irq:      Interrupt number that has fired
+ * @cookie:   PDC state for DMA engine that generated the interrupt
+ *
+ * We have to clear the device interrupt status flags here. So cache the
+ * status for later use in the thread function. Other than that, just return
+ * WAKE_THREAD to invoke the thread function.
+ *
+ * Return: IRQ_WAKE_THREAD if interrupt is ours
+ *         IRQ_NONE otherwise
+ */
+static irqreturn_t pdc_irq_handler(int irq, void *cookie)
+{
+	struct pdc_state *pdcs = cookie;
+	u32 intstatus = ioread32(pdcs->pdc_reg_vbase + PDC_INTSTATUS_OFFSET);
+
+	if (intstatus & PDC_XMTINTEN_0)
+		set_bit(PDC_XMTINT_0, &pdcs->intstatus);
+	if (intstatus & PDC_RCVINTEN_0)
+		set_bit(PDC_RCVINT_0, &pdcs->intstatus);
+
+	/* Clear interrupt flags in device */
+	iowrite32(intstatus, pdcs->pdc_reg_vbase + PDC_INTSTATUS_OFFSET);
+
+	/* Wakeup IRQ thread */
+	if (pdcs && (irq == pdcs->pdc_irq) && (intstatus & PDC_INTMASK))
+		return IRQ_WAKE_THREAD;
+
+	return IRQ_NONE;
+}
+
+/**
+ * pdc_irq_thread() - Function invoked on deferred thread when a DMA tx has
+ * completed or data is available to receive.
+ * @irq:    Interrupt number
+ * @cookie: PDC state for PDC that generated the interrupt
+ *
+ * On DMA tx complete, notify the mailbox client. On DMA rx complete, process
+ * as many SPU response messages as are available and send each to the mailbox
+ * client.
+ *
+ * Return: IRQ_HANDLED if we recognized and handled the interrupt
+ *         IRQ_NONE otherwise
+ */
+static irqreturn_t pdc_irq_thread(int irq, void *cookie)
+{
+	struct pdc_state *pdcs = cookie;
+	struct mbox_controller *mbc;
+	struct mbox_chan *chan;
+	bool tx_int;
+	bool rx_int;
+	int rx_status;
+	struct brcm_message mssg;
+
+	tx_int = test_and_clear_bit(PDC_XMTINT_0, &pdcs->intstatus);
+	rx_int = test_and_clear_bit(PDC_RCVINT_0, &pdcs->intstatus);
+
+	if (pdcs && (tx_int || rx_int)) {
+		dev_dbg(&pdcs->pdev->dev,
+			"%s() got irq %d with tx_int %s, rx_int %s",
+			__func__, irq,
+			tx_int ? "set" : "clear", rx_int ? "set" : "clear");
+
+		mbc = &pdcs->mbc;
+		chan = &mbc->chans[0];
+
+		if (tx_int) {
+			dev_dbg(&pdcs->pdev->dev, "%s(): tx done", __func__);
+			/* only one frame in flight at a time */
+			mbox_chan_txdone(chan, PDC_SUCCESS);
+		}
+		if (rx_int) {
+			while (1) {
+				/* Could be many frames ready */
+				memset(&mssg, 0, sizeof(mssg));
+				mssg.type = BRCM_MESSAGE_SPU;
+				rx_status = pdc_receive(pdcs, &mssg);
+				if (rx_status >= 0) {
+					dev_dbg(&pdcs->pdev->dev,
+						"%s(): invoking client rx cb",
+						__func__);
+					mbox_chan_received_data(chan, &mssg);
+				} else {
+					dev_dbg(&pdcs->pdev->dev,
+						"%s(): no SPU response available",
+						__func__);
+					break;
+				}
+			}
+		}
+		return IRQ_HANDLED;
+	}
+	return IRQ_NONE;
+}
+
+/**
+ * pdc_ring_init() - Allocate DMA rings and initialize constant fields of
+ * descriptors in one ringset.
+ * @pdcs:    PDC instance state
+ * @ringset: index of ringset being used
+ *
+ * Return: PDC_SUCCESS if ring initialized
+ *         < 0 otherwise
+ */
+static int pdc_ring_init(struct pdc_state *pdcs, int ringset)
+{
+	int i;
+	int err = PDC_SUCCESS;
+	struct dma64 *dma_reg;
+	struct device *dev = &pdcs->pdev->dev;
+	struct pdc_ring_alloc tx;
+	struct pdc_ring_alloc rx;
+
+	/* Allocate tx ring */
+	tx.vbase = dma_pool_zalloc(pdcs->ring_pool, GFP_KERNEL, &tx.dmabase);
+	if (!tx.vbase) {
+		err = -ENOMEM;
+		goto done;
+	}
+
+	/* Allocate rx ring */
+	rx.vbase = dma_pool_zalloc(pdcs->ring_pool, GFP_KERNEL, &rx.dmabase);
+	if (!rx.vbase) {
+		err = -ENOMEM;
+		goto fail_dealloc;
+	}
+
+	dev_dbg(dev, " - base DMA addr of tx ring      %#llx", tx.dmabase);
+	dev_dbg(dev, " - base virtual addr of tx ring  %p", tx.vbase);
+	dev_dbg(dev, " - base DMA addr of rx ring      %#llx", rx.dmabase);
+	dev_dbg(dev, " - base virtual addr of rx ring  %p", rx.vbase);
+
+	/* lock after ring allocation to avoid scheduling while atomic */
+	spin_lock(&pdcs->pdc_lock);
+
+	memcpy(&pdcs->tx_ring_alloc, &tx, sizeof(tx));
+	memcpy(&pdcs->rx_ring_alloc, &rx, sizeof(rx));
+
+	pdcs->rxin = 0;
+	pdcs->rx_msg_start = 0;
+	pdcs->last_rx_curr = 0;
+	pdcs->rxout = 0;
+	pdcs->txin = 0;
+	pdcs->tx_msg_start = 0;
+	pdcs->txout = 0;
+
+	/* Set descriptor array base addresses */
+	pdcs->txd_64 = (struct dma64dd *)pdcs->tx_ring_alloc.vbase;
+	pdcs->rxd_64 = (struct dma64dd *)pdcs->rx_ring_alloc.vbase;
+
+	/* Tell device the base DMA address of each ring */
+	dma_reg = &pdcs->regs->dmaregs[ringset];
+	iowrite32(lower_32_bits(pdcs->tx_ring_alloc.dmabase),
+		  (void *)&dma_reg->dmaxmt.addrlow);
+	iowrite32(upper_32_bits(pdcs->tx_ring_alloc.dmabase),
+		  (void *)&dma_reg->dmaxmt.addrhigh);
+
+	iowrite32(lower_32_bits(pdcs->rx_ring_alloc.dmabase),
+		  (void *)&dma_reg->dmarcv.addrlow);
+	iowrite32(upper_32_bits(pdcs->rx_ring_alloc.dmabase),
+		  (void *)&dma_reg->dmarcv.addrhigh);
+
+	/* Initialize descriptors */
+	for (i = 0; i < PDC_RING_ENTRIES; i++) {
+		/* Every tx descriptor can be used for start of frame. */
+		if (i != pdcs->ntxpost) {
+			iowrite32(D64_CTRL1_SOF | D64_CTRL1_EOF,
+				  (void *)&pdcs->txd_64[i].ctrl1);
+		} else {
+			/* Last descriptor in ringset. Set End of Table. */
+			iowrite32(D64_CTRL1_SOF | D64_CTRL1_EOF |
+				  D64_CTRL1_EOT,
+				  (void *)&pdcs->txd_64[i].ctrl1);
+		}
+
+		/* Every rx descriptor can be used for start of frame */
+		if (i != pdcs->nrxpost) {
+			iowrite32(D64_CTRL1_SOF,
+				  (void *)&pdcs->rxd_64[i].ctrl1);
+		} else {
+			/* Last descriptor in ringset. Set End of Table. */
+			iowrite32(D64_CTRL1_SOF | D64_CTRL1_EOT,
+				  (void *)&pdcs->rxd_64[i].ctrl1);
+		}
+	}
+	spin_unlock(&pdcs->pdc_lock);
+	return PDC_SUCCESS;
+
+fail_dealloc:
+	dma_pool_free(pdcs->ring_pool, tx.vbase, tx.dmabase);
+done:
+	return err;
+}
+
+static void pdc_ring_free(struct pdc_state *pdcs)
+{
+	if (pdcs->tx_ring_alloc.vbase) {
+		dma_pool_free(pdcs->ring_pool, pdcs->tx_ring_alloc.vbase,
+			      pdcs->tx_ring_alloc.dmabase);
+		pdcs->tx_ring_alloc.vbase = NULL;
+	}
+
+	if (pdcs->rx_ring_alloc.vbase) {
+		dma_pool_free(pdcs->ring_pool, pdcs->rx_ring_alloc.vbase,
+			      pdcs->rx_ring_alloc.dmabase);
+		pdcs->rx_ring_alloc.vbase = NULL;
+	}
+}
+
+/**
+ * pdc_send_data() - mailbox send_data function
+ * @chan:	The mailbox channel on which the data is sent. The channel
+ *              corresponds to a DMA ringset.
+ * @data:	The mailbox message to be sent. The message must be a
+ *              brcm_message structure.
+ *
+ * This function is registered as the send_data function for the mailbox
+ * controller. From the destination scatterlist in the mailbox message, it
+ * creates a sequence of receive descriptors in the rx ring. From the source
+ * scatterlist, it creates a sequence of transmit descriptors in the tx ring.
+ * After creating the descriptors, it writes the rx ptr and tx ptr registers to
+ * initiate the DMA transfer.
+ *
+ * This function does the DMA map and unmap of the src and dst scatterlists in
+ * the mailbox message.
+ *
+ * Return: 0 if successful
+ *	   -ENOTSUPP if the mailbox message is a type this driver does not
+ *			support
+ *         < 0 if an error
+ */
+static int pdc_send_data(struct mbox_chan *chan, void *data)
+{
+	struct pdc_state *pdcs = chan->con_priv;
+	struct device *dev = &pdcs->pdev->dev;
+	struct brcm_message *mssg = data;
+	int err = PDC_SUCCESS;
+	int src_nent;
+	int dst_nent;
+	int nent;
+
+	if (mssg->type != BRCM_MESSAGE_SPU)
+		return -ENOTSUPP;
+
+	src_nent = sg_nents(mssg->spu.src);
+	if (src_nent) {
+		nent = dma_map_sg(dev, mssg->spu.src, src_nent, DMA_TO_DEVICE);
+		if (nent == 0)
+			return -EIO;
+	}
+
+	dst_nent = sg_nents(mssg->spu.dst);
+	if (dst_nent) {
+		nent = dma_map_sg(dev, mssg->spu.dst, dst_nent,
+				  DMA_FROM_DEVICE);
+		if (nent == 0) {
+			dma_unmap_sg(dev, mssg->spu.src, src_nent,
+				     DMA_TO_DEVICE);
+			return -EIO;
+		}
+	}
+
+	spin_lock(&pdcs->pdc_lock);
+
+	/* Create rx descriptors to SPU catch response */
+	err = pdc_rx_list_init(pdcs, mssg->spu.dst, mssg->ctx);
+	err |= pdc_rx_list_sg_add(pdcs, mssg->spu.dst);
+
+	/* Create tx descriptors to submit SPU request */
+	err |= pdc_tx_list_sg_add(pdcs, mssg->spu.src);
+	err |= pdc_tx_list_final(pdcs);	/* initiate transfer */
+
+	spin_unlock(&pdcs->pdc_lock);
+
+	if (err)
+		dev_err(&pdcs->pdev->dev,
+			"%s failed with error %d", __func__, err);
+
+	return err;
+}
+
+static int pdc_startup(struct mbox_chan *chan)
+{
+	return pdc_ring_init(chan->con_priv, PDC_RINGSET);
+}
+
+static void pdc_shutdown(struct mbox_chan *chan)
+{
+	struct pdc_state *pdcs = chan->con_priv;
+
+	if (pdcs)
+		dev_dbg(&pdcs->pdev->dev,
+			"Shutdown mailbox channel for PDC %u", pdcs->pdc_idx);
+
+	pdc_ring_free(pdcs);
+}
+
+/**
+ * pdc_hw_init() - Use the given initialization parameters to initialize the
+ * state for one of the PDCs.
+ * @dev:   device structure for PDC
+ * @pdcs:  state of the PDC
+ * @parms: parameter values to set
+ */
+static
+void pdc_hw_init(struct device *dev, struct pdc_state *pdcs,
+		 struct hw_init_parms *parms)
+{
+	struct dma64 *dma_reg;
+
+	int ringset = PDC_RINGSET;
+
+	dev_dbg(dev, "PDC %u initial values:", pdcs->pdc_idx);
+	dev_dbg(dev, "state structure:                   %p",
+		pdcs);
+	dev_dbg(dev, " - base physical addr of hw regs   %#llx",
+		parms->hw_pbase);
+	dev_dbg(dev, " - base virtual addr of hw regs    %p",
+		parms->hw_vbase);
+
+	/* initialize data structures */
+	pdcs->regs = (struct pdc_regs *)parms->hw_vbase;
+	pdcs->txregs_64 = (struct dma64_regs *)
+	    (void *)(((u8 *)parms->hw_vbase) +
+		     PDC_TXREGS_OFFSET + (sizeof(struct dma64) * ringset));
+	pdcs->rxregs_64 = (struct dma64_regs *)
+	    (void *)(((u8 *)parms->hw_vbase) +
+		     PDC_RXREGS_OFFSET + (sizeof(struct dma64) * ringset));
+
+	pdcs->ntxd = parms->ring_entries;
+	pdcs->nrxd = parms->ring_entries;
+	pdcs->ntxpost = parms->ring_entries - 1;
+	pdcs->nrxpost = parms->ring_entries - 1;
+	pdcs->regs->intmask = 0;
+
+	dma_reg = &pdcs->regs->dmaregs[ringset];
+	iowrite32(0, (void *)&dma_reg->dmaxmt.ptr);
+	iowrite32(0, (void *)&dma_reg->dmarcv.ptr);
+
+	iowrite32(PDC_TX_CTL, (void *)&dma_reg->dmaxmt.control);
+
+	iowrite32(PDC_RX_CTL + (pdcs->rx_status_len << 1),
+		  (void *)&dma_reg->dmarcv.control);
+
+	if (pdcs->pdc_resp_hdr_len == PDC_SPU2_RESP_HDR_LEN)
+		iowrite32(PDC_CKSUM_CTRL,
+			  pdcs->pdc_reg_vbase + PDC_CKSUM_CTRL_OFFSET);
+}
+
+/**
+ * pdc_rx_buf_pool_create() - Pool of receive buffers used to catch the metadata
+ * header returned with each response message.
+ * @dev:  device structure
+ * @pdcs: PDC state structure
+ *
+ * The metadata is not returned to the mailbox client. So the PDC driver
+ * manages these buffers.
+ *
+ * Return: PDC_SUCCESS
+ *         -ENOMEM if pool creation fails
+ */
+static int pdc_rx_buf_pool_create(struct device *dev, struct pdc_state *pdcs)
+{
+	pdcs->pdc_resp_hdr_len = pdcs->rx_status_len;
+	if (pdcs->use_bcm_hdr)
+		pdcs->pdc_resp_hdr_len += BCM_HDR_LEN;
+
+	pdcs->rx_buf_pool = dma_pool_create("pdc rx bufs", dev,
+					    pdcs->pdc_resp_hdr_len,
+					    RX_BUF_ALIGN, 0);
+	if (!pdcs->rx_buf_pool)
+		return -ENOMEM;
+
+	return PDC_SUCCESS;
+}
+
+/**
+ * pdc_interrupts_init() - Initialize the interrupt configuration for a PDC and
+ * specify a threaded IRQ handler for deferred handling of interrupts outside of
+ * interrupt context.
+ * @pdev:   platform device
+ * @pdcs:   PDC state
+ *
+ * Set the interrupt mask for transmit and receive done.
+ * Set the lazy interrupt frame count to generate an interrupt for just one pkt.
+ *
+ * Return:  PDC_SUCCESS
+ *          <0 if threaded irq request fails
+ */
+static int pdc_interrupts_init(struct platform_device *pdev,
+			       struct pdc_state *pdcs)
+{
+	struct device *dev = &pdev->dev;
+	struct device_node *dn = pdev->dev.of_node;
+	int err;
+
+	pdcs->intstatus = 0;
+
+	/* interrupt configuration */
+	iowrite32(PDC_INTMASK, pdcs->pdc_reg_vbase + PDC_INTMASK_OFFSET);
+	iowrite32(PDC_LAZY_INT, pdcs->pdc_reg_vbase + PDC_RCVLAZY0_OFFSET);
+
+	/* read irq from device tree */
+	pdcs->pdc_irq = irq_of_parse_and_map(dn, 0);
+	dev_dbg(dev, "pdc device %s irq %u for pdcs %p",
+		dev_name(dev), pdcs->pdc_irq, pdcs);
+	err = devm_request_threaded_irq(dev, pdcs->pdc_irq,
+					pdc_irq_handler,
+					pdc_irq_thread, 0, dev_name(dev), pdcs);
+	if (err) {
+		dev_err(dev, "threaded tx IRQ %u request failed with err %d\n",
+			pdcs->pdc_irq, err);
+		return err;
+	}
+	return PDC_SUCCESS;
+}
+
+static const struct mbox_chan_ops pdc_mbox_chan_ops = {
+	.send_data = pdc_send_data,
+	.startup = pdc_startup,
+	.shutdown = pdc_shutdown
+};
+
+/**
+ * pdc_mb_init() - Initialize the mailbox controller.
+ * @pdcs:  PDC state
+ *
+ * Each PDC is a mailbox controller. Each ringset is a mailbox channel. Kernel
+ * driver only uses one ringset and thus one mb channel. PDC uses the transmit
+ * complete interrupt to determine when a mailbox message has successfully been
+ * transmitted.
+ *
+ * Return: 0 on success
+ *         < 0 if there is an allocation or registration failure
+ */
+static int pdc_mb_init(struct pdc_state *pdcs)
+{
+	struct device *dev = &pdcs->pdev->dev;
+	struct mbox_controller *mbc;
+	int chan_index;
+	int err;
+
+	mbc = &pdcs->mbc;
+	mbc->dev = dev;
+	mbc->ops = &pdc_mbox_chan_ops;
+	mbc->num_chans = 1;
+	mbc->chans = devm_kcalloc(dev, mbc->num_chans, sizeof(*mbc->chans),
+				  GFP_KERNEL);
+	if (!mbc->chans)
+		return -ENOMEM;
+
+	mbc->txdone_irq = true;
+	mbc->txdone_poll = false;
+	for (chan_index = 0; chan_index < mbc->num_chans; chan_index++)
+		mbc->chans[chan_index].con_priv = pdcs;
+
+	/* Register mailbox controller */
+	err = mbox_controller_register(mbc);
+	if (err) {
+		dev_crit(dev,
+			 "Failed to register PDC mailbox controller. Error %d.",
+			 err);
+		return err;
+	}
+	return 0;
+}
+
+/**
+ * pdc_dt_read() - Read application-specific data from device tree.
+ * @pdev:  Platform device
+ * @pdcs:  PDC state
+ *
+ * Reads the number of bytes of receive status that precede each received frame.
+ * Reads whether transmit and received frames should be preceded by an 8-byte
+ * BCM header.
+ *
+ * Return: 0 if successful
+ *         -ENODEV if device not available
+ */
+static int pdc_dt_read(struct platform_device *pdev, struct pdc_state *pdcs)
+{
+	struct device *dev = &pdev->dev;
+	struct device_node *dn = pdev->dev.of_node;
+	int err;
+
+	err = of_property_read_u32(dn, "brcm,rx-status-len",
+				   &pdcs->rx_status_len);
+	if (err < 0)
+		dev_err(dev,
+			"%s failed to get DMA receive status length from device tree",
+			__func__);
+
+	pdcs->use_bcm_hdr = of_property_read_bool(dn, "brcm,use-bcm-hdr");
+
+	return 0;
+}
+
+/**
+ * pdc_probe() - Probe function for PDC driver.
+ * @pdev:   PDC platform device
+ *
+ * Reserve and map register regions defined in device tree.
+ * Allocate and initialize tx and rx DMA rings.
+ * Initialize a mailbox controller for each PDC.
+ *
+ * Return: 0 if successful
+ *         < 0 if an error
+ */
+static int pdc_probe(struct platform_device *pdev)
+{
+	int err = 0;
+	struct device *dev = &pdev->dev;
+	struct resource *pdc_regs;
+	struct pdc_state *pdcs;
+	struct hw_init_parms hw_parms;	/* params for initializing spu-dma */
+
+	/* PDC state for one SPU */
+	pdcs = devm_kzalloc(dev, sizeof(*pdcs), GFP_KERNEL);
+	if (!pdcs) {
+		err = -ENOMEM;
+		goto cleanup;
+	}
+
+	spin_lock_init(&pdcs->pdc_lock);
+	pdcs->pdev = pdev;
+	platform_set_drvdata(pdev, pdcs);
+	pdcs->pdc_idx = pdcg.num_spu;
+	pdcg.num_spu++;
+
+	err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32));
+	if (err) {
+		dev_warn(dev, "PDC device cannot perform DMA. Error %d.", err);
+		goto cleanup;
+	}
+
+	/* Create DMA pool for tx ring */
+	pdcs->ring_pool = dma_pool_create("pdc rings", dev, PDC_RING_SIZE,
+					  RING_ALIGN, 0);
+	if (!pdcs->ring_pool) {
+		err = -ENOMEM;
+		goto cleanup;
+	}
+
+	err = pdc_dt_read(pdev, pdcs);
+	if (err)
+		goto cleanup_ring_pool;
+
+	pdc_regs = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	if (!pdc_regs) {
+		err = -ENODEV;
+		goto cleanup_ring_pool;
+	}
+	dev_dbg(dev, "PDC register region res.start = %#llx, res.end = %#llx",
+		pdc_regs->start, pdc_regs->end);
+
+	pdcs->pdc_reg_vbase = devm_ioremap_resource(&pdev->dev, pdc_regs);
+	if (IS_ERR(pdcs->pdc_reg_vbase)) {
+		err = PTR_ERR(pdcs->pdc_reg_vbase);
+		dev_err(&pdev->dev, "Failed to map registers: %d\n", err);
+		goto cleanup_ring_pool;
+	}
+
+	/* create rx buffer pool after dt read to know how big buffers are */
+	err = pdc_rx_buf_pool_create(dev, pdcs);
+	if (err)
+		goto cleanup_ring_pool;
+
+	hw_parms.hw_pbase = (dma_addr_t)(pdc_regs->start);
+	hw_parms.hw_vbase = pdcs->pdc_reg_vbase;
+
+	hw_parms.ring_entries = PDC_RING_ENTRIES;
+
+	pdc_hw_init(dev, pdcs, &hw_parms);
+
+	err = pdc_interrupts_init(pdev, pdcs);
+	if (err)
+		goto cleanup_buf_pool;
+
+	/* Initialize mailbox controller */
+	err = pdc_mb_init(pdcs);
+	if (err)
+		goto cleanup_buf_pool;
+
+	pdcs->debugfs_stats = NULL;
+	pdc_setup_debugfs(pdcs);
+
+	dev_dbg(dev, "pdc_probe() successful");
+	return PDC_SUCCESS;
+
+cleanup_buf_pool:
+	dma_pool_destroy(pdcs->rx_buf_pool);
+
+cleanup_ring_pool:
+	dma_pool_destroy(pdcs->ring_pool);
+
+cleanup:
+	return err;
+}
+
+static int pdc_remove(struct platform_device *pdev)
+{
+	struct pdc_state *pdcs = platform_get_drvdata(pdev);
+
+	pdc_free_debugfs();
+
+	mbox_controller_unregister(&pdcs->mbc);
+
+	dma_pool_destroy(pdcs->rx_buf_pool);
+	dma_pool_destroy(pdcs->ring_pool);
+	return 0;
+}
+
+static const struct of_device_id pdc_mbox_of_match[] = {
+	{.compatible = "brcm,iproc-pdc-mbox"},
+	{ /* sentinel */ }
+};
+MODULE_DEVICE_TABLE(of, pdc_mbox_of_match);
+
+static struct platform_driver pdc_mbox_driver = {
+	.probe = pdc_probe,
+	.remove = pdc_remove,
+	.driver = {
+		   .name = "brcm-iproc-pdc-mbox",
+		   .of_match_table = of_match_ptr(pdc_mbox_of_match),
+		   },
+};
+module_platform_driver(pdc_mbox_driver);
+
+MODULE_AUTHOR("Rob Rice <rob.rice@broadcom.com>");
+MODULE_DESCRIPTION("Broadcom PDC mailbox driver");
+MODULE_LICENSE("GPL v2");
diff --git a/drivers/mailbox/mailbox-pdc/pdc.h b/drivers/mailbox/mailbox-pdc/pdc.h
new file mode 100644
index 0000000..17e1464
--- /dev/null
+++ b/drivers/mailbox/mailbox-pdc/pdc.h
@@ -0,0 +1,303 @@
+/*
+ * Copyright 2016 Broadcom
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, version 2, as
+ * published by the Free Software Foundation (the "GPL").
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License version 2 (GPLv2) for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * version 2 (GPLv2) along with this source code.
+ */
+#ifndef PDC_H
+#define PDC_H
+
+#include <linux/scatterlist.h>
+#include <linux/mailbox_controller.h>
+
+#define PDC_SUCCESS  0
+
+#define RING_ENTRY_SIZE   sizeof(struct dma64dd)
+
+/* # entries in PDC dma ring */
+#define PDC_RING_ENTRIES  128
+#define PDC_RING_SIZE    (PDC_RING_ENTRIES * RING_ENTRY_SIZE)
+/* Rings are 8k aligned */
+#define RING_ALIGN_ORDER  13
+#define RING_ALIGN        BIT(RING_ALIGN_ORDER)
+
+#define RX_BUF_ALIGN_ORDER  5
+#define RX_BUF_ALIGN	    BIT(RX_BUF_ALIGN_ORDER)
+
+/* descriptor bumping macros */
+#define XXD(x, max_mask)              ((x) & (max_mask))
+#define TXD(x, max_mask)              XXD((x), (max_mask))
+#define RXD(x, max_mask)              XXD((x), (max_mask))
+#define NEXTTXD(i, max_mask)          TXD((i) + 1, (max_mask))
+#define PREVTXD(i, max_mask)          TXD((i) - 1, (max_mask))
+#define NEXTRXD(i, max_mask)          RXD((i) + 1, (max_mask))
+#define PREVRXD(i, max_mask)          RXD((i) - 1, (max_mask))
+#define NTXDACTIVE(h, t, max_mask)    TXD((t) - (h), (max_mask))
+#define NRXDACTIVE(h, t, max_mask)    RXD((t) - (h), (max_mask))
+
+struct pdc_dma_map {
+	void *ctx;          /* opaque context associated with frame */
+};
+
+struct hw_init_parms {
+	dma_addr_t hw_pbase;         /* base physical addr of PDC regs */
+	void *hw_vbase;              /* base virtual addr of PDC regs */
+
+	u32  ring_entries;           /* number of ring entries */
+};
+
+/* dma descriptor */
+struct dma64dd {
+	u32 ctrl1;      /* misc control bits */
+	u32 ctrl2;      /* buffer count and address extension */
+	u32 addrlow;    /* memory address of the date buffer, bits 31:0 */
+	u32 addrhigh;   /* memory address of the date buffer, bits 63:32 */
+};
+
+/* dma registers per channel(xmt or rcv) */
+struct dma64_regs {
+	u32  control;   /* enable, et al */
+	u32  ptr;       /* last descriptor posted to chip */
+	u32  addrlow;   /* descriptor ring base address low 32-bits */
+	u32  addrhigh;  /* descriptor ring base address bits 63:32 */
+	u32  status0;   /* last rx descriptor written by hw */
+	u32  status1;   /* driver does not use */
+};
+
+/* cpp contortions to concatenate w/arg prescan */
+#ifndef PAD
+#define _PADLINE(line)  pad ## line
+#define _XSTR(line)     _PADLINE(line)
+#define PAD             _XSTR(__LINE__)
+#endif  /* PAD */
+
+/* dma registers. matches hw layout. */
+struct dma64 {
+	struct dma64_regs dmaxmt;  /* dma tx */
+	u32          PAD[2];
+	struct dma64_regs dmarcv;  /* dma rx */
+	u32          PAD[2];
+};
+
+/* PDC registers */
+struct pdc_regs {
+	u32  devcontrol;             /* 0x000 */
+	u32  devstatus;              /* 0x004 */
+	u32  PAD;
+	u32  biststatus;             /* 0x00c */
+	u32  PAD[4];
+	u32  intstatus;              /* 0x020 */
+	u32  intmask;                /* 0x024 */
+	u32  gptimer;                /* 0x028 */
+
+	u32  PAD;
+	u32  intrcvlazy_0;           /* 0x030 */
+	u32  intrcvlazy_1;           /* 0x034 */
+	u32  intrcvlazy_2;           /* 0x038 */
+	u32  intrcvlazy_3;           /* 0x03c */
+
+	u32  PAD[48];
+	u32  removed_intrecvlazy;    /* 0x100 */
+	u32  flowctlthresh;          /* 0x104 */
+	u32  wrrthresh;              /* 0x108 */
+	u32  gmac_idle_cnt_thresh;   /* 0x10c */
+
+	u32  PAD[4];
+	u32  ifioaccessaddr;         /* 0x120 */
+	u32  ifioaccessbyte;         /* 0x124 */
+	u32  ifioaccessdata;         /* 0x128 */
+
+	u32  PAD[21];
+	u32  phyaccess;              /* 0x180 */
+	u32  PAD;
+	u32  phycontrol;             /* 0x188 */
+	u32  txqctl;                 /* 0x18c */
+	u32  rxqctl;                 /* 0x190 */
+	u32  gpioselect;             /* 0x194 */
+	u32  gpio_output_en;         /* 0x198 */
+	u32  PAD;                    /* 0x19c */
+	u32  txq_rxq_mem_ctl;        /* 0x1a0 */
+	u32  memory_ecc_status;      /* 0x1a4 */
+	u32  serdes_ctl;             /* 0x1a8 */
+	u32  serdes_status0;         /* 0x1ac */
+	u32  serdes_status1;         /* 0x1b0 */
+	u32  PAD[11];                /* 0x1b4-1dc */
+	u32  clk_ctl_st;             /* 0x1e0 */
+	u32  hw_war;                 /* 0x1e4 */
+	u32  pwrctl;                 /* 0x1e8 */
+	u32  PAD[5];
+
+#define PDC_NUM_DMA_RINGS   4
+	struct dma64 dmaregs[PDC_NUM_DMA_RINGS];  /* 0x0200 - 0x2fc */
+
+	/* more registers follow, but we don't use them */
+};
+
+/* structure for allocating/freeing DMA rings */
+struct pdc_ring_alloc {
+	dma_addr_t  dmabase; /* DMA address of start of ring */
+	void	   *vbase;   /* base kernel virtual address of ring */
+	u32	    size;    /* ring allocation size in bytes */
+};
+
+/* PDC state structure */
+struct pdc_state {
+	/* synchronize access to this PDC state structure */
+	spinlock_t pdc_lock;
+
+	/* Index of the PDC whose state is in this structure instance */
+	u8 pdc_idx;
+
+	/* Platform device for this PDC instance */
+	struct platform_device *pdev;
+
+	/*
+	 * Each PDC instance has a mailbox controller. PDC receives request
+	 * messages through mailboxes, and sends response messages through the
+	 * mailbox framework.
+	 */
+	struct mbox_controller mbc;
+
+	unsigned int pdc_irq;
+
+	/*
+	 * Last interrupt status read from PDC device. Saved in interrupt
+	 * handler so the handler can clear the interrupt in the device,
+	 * and the interrupt thread called later can know which interrupt
+	 * bits are active.
+	 */
+	unsigned long intstatus;
+
+	/* Number of bytes of receive status prior to each rx frame */
+	u32 rx_status_len;
+	/* Whether a BCM header is prepended to each frame */
+	bool use_bcm_hdr;
+	/* Sum of length of BCM header and rx status header */
+	u32 pdc_resp_hdr_len;
+
+	/* The base virtual address of DMA hw registers */
+	void __iomem *pdc_reg_vbase;
+
+	/* Pool for allocation of DMA rings */
+	struct dma_pool *ring_pool;
+
+	/* Pool for allocation of metadata buffers for response messages */
+	struct dma_pool *rx_buf_pool;
+
+	/*
+	 * The base virtual address of DMA tx/rx descriptor rings. Corresponding
+	 * DMA address and size of ring allocation.
+	 */
+	struct pdc_ring_alloc tx_ring_alloc;
+	struct pdc_ring_alloc rx_ring_alloc;
+
+	struct pdc_regs *regs;    /* start of PDC registers */
+
+	struct dma64_regs *txregs_64; /* dma tx engine registers */
+	struct dma64_regs *rxregs_64; /* dma rx engine registers */
+
+	/*
+	 * Arrays of PDC_RING_ENTRIES descriptors
+	 * To use multiple ringsets, this needs to be extended
+	 */
+	struct dma64dd   *txd_64;  /* tx descriptor ring */
+	struct dma64dd   *rxd_64;  /* rx descriptor ring */
+
+	/* descriptor ring sizes */
+	u32      ntxd;       /* # tx descriptors */
+	u32      nrxd;       /* # rx descriptors */
+	u32      nrxpost;    /* # rx buffers to keep posted */
+	u32      ntxpost;    /* max number of tx buffers that can be posted */
+
+	/*
+	 * Index of next tx descriptor to reclaim. That is, the descriptor
+	 * index of the oldest tx buffer for which the host has yet to process
+	 * the corresponding response.
+	 */
+	u32  txin;
+
+	/*
+	 * Index of the first receive descriptor for the sequence of
+	 * message fragments currently under construction. Used to build up
+	 * the rxin_numd count for a message. Updated to rxout when the host
+	 * starts a new sequence of rx buffers for a new message.
+	 */
+	u32  tx_msg_start;
+
+	/* Index of next tx descriptor to post. */
+	u32  txout;
+
+	/*
+	 * Number of tx descriptors associated with the message that starts
+	 * at this tx descriptor index.
+	 */
+	u32      txin_numd[PDC_RING_ENTRIES];
+
+	/*
+	 * Index of next rx descriptor to reclaim. This is the index of
+	 * the next descriptor whose data has yet to be processed by the host.
+	 */
+	u32  rxin;
+
+	/*
+	 * Index of the first receive descriptor for the sequence of
+	 * message fragments currently under construction. Used to build up
+	 * the rxin_numd count for a message. Updated to rxout when the host
+	 * starts a new sequence of rx buffers for a new message.
+	 */
+	u32  rx_msg_start;
+
+	/*
+	 * Saved value of current hardware rx descriptor index.
+	 * The last rx buffer written by the hw is the index previous to
+	 * this one.
+	 */
+	u32  last_rx_curr;
+
+	/* Index of next rx descriptor to post. */
+	u32  rxout;
+
+	/*
+	 * opaque context associated with frame that starts at each
+	 * rx ring index.
+	 */
+	void *rxp_ctx[PDC_RING_ENTRIES];
+
+	/*
+	 * Scatterlists used to form request and reply frames beginning at a
+	 * given ring index. Retained in order to unmap each sg after reply
+	 * is processed
+	 */
+	struct scatterlist *src_sg[PDC_RING_ENTRIES];
+	struct scatterlist *dst_sg[PDC_RING_ENTRIES];
+
+	/*
+	 * Number of rx descriptors associated with the message that starts
+	 * at this descriptor index. Not set for every index. For example,
+	 * if descriptor index i points to a scatterlist with 4 entries, then
+	 * the next three descriptor indexes don't have a value set.
+	 */
+	u32  rxin_numd[PDC_RING_ENTRIES];
+
+	void *resp_hdr[PDC_RING_ENTRIES];
+	dma_addr_t resp_hdr_daddr[PDC_RING_ENTRIES];
+
+	struct dentry *debugfs_stats;  /* debug FS stats file for this PDC */
+
+	/* counters */
+	u32  pdc_requests;    /* number of request messages submitted */
+	u32  pdc_replies;     /* number of reply messages received */
+	u32  txnobuf;         /* count of tx ring full */
+	u32  rxnobuf;         /* count of rx ring full */
+	u32  rx_oflow;        /* count of rx overflows */
+};
+#endif
diff --git a/drivers/mailbox/mailbox-pdc/pdc_debug.c b/drivers/mailbox/mailbox-pdc/pdc_debug.c
new file mode 100644
index 0000000..891a937
--- /dev/null
+++ b/drivers/mailbox/mailbox-pdc/pdc_debug.c
@@ -0,0 +1,101 @@
+/*
+ * Copyright 2016 Broadcom
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, version 2, as
+ * published by the Free Software Foundation (the "GPL").
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License version 2 (GPLv2) for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * version 2 (GPLv2) along with this source code.
+ */
+
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/debugfs.h>
+
+#include "pdc.h"
+
+/* top level debug FS directory for PDC driver */
+static struct dentry *debugfs_dir;
+
+static ssize_t pdc_debugfs_read(struct file *filp, char __user *ubuf,
+				size_t count, loff_t *offp)
+{
+	struct pdc_state *pdcs;
+	char *buf;
+	ssize_t ret, out_offset, out_count;
+
+	out_count = 512;
+
+	buf = kmalloc(out_count, GFP_KERNEL);
+	if (!buf)
+		return -ENOMEM;
+
+	pdcs = filp->private_data;
+	out_offset = 0;
+	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+			       "SPU %u stats:\n", pdcs->pdc_idx);
+	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+			       "PDC requests............%u\n",
+			       pdcs->pdc_requests);
+	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+			       "PDC responses...........%u\n",
+			       pdcs->pdc_replies);
+	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+			       "Tx err ring full........%u\n",
+			       pdcs->txnobuf);
+	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+			       "Rx err ring full........%u\n",
+			       pdcs->rxnobuf);
+	out_offset += snprintf(buf + out_offset, out_count - out_offset,
+			       "Receive overflow........%u\n",
+			       pdcs->rx_oflow);
+
+	if (out_offset > out_count)
+		out_offset = out_count;
+
+	ret = simple_read_from_buffer(ubuf, count, offp, buf, out_offset);
+	kfree(buf);
+	return ret;
+}
+
+static const struct file_operations pdc_debugfs_stats = {
+	.owner = THIS_MODULE,
+	.open = simple_open,
+	.read = pdc_debugfs_read,
+};
+
+/**
+ * pdc_setup_debugfs() - Create the debug FS directories. If the top-level
+ * directory has not yet been created, create it now. Create a stats file in
+ * this directory for a SPU.
+ * @pdcs: PDC state structure
+ */
+void pdc_setup_debugfs(struct pdc_state *pdcs)
+{
+	char spu_stats_name[16];
+
+	if (!debugfs_initialized())
+		return;
+
+	snprintf(spu_stats_name, 16, "pdc%d_stats", pdcs->pdc_idx);
+	if (!debugfs_dir)
+		debugfs_dir = debugfs_create_dir(KBUILD_MODNAME, NULL);
+
+	pdcs->debugfs_stats = debugfs_create_file(spu_stats_name, S_IRUSR,
+						  debugfs_dir, pdcs,
+						  &pdc_debugfs_stats);
+}
+
+void pdc_free_debugfs(void)
+{
+	if (debugfs_dir && simple_empty(debugfs_dir)) {
+		debugfs_remove_recursive(debugfs_dir);
+		debugfs_dir = NULL;
+	}
+}
diff --git a/drivers/mailbox/mailbox-pdc/pdc_debug.h b/drivers/mailbox/mailbox-pdc/pdc_debug.h
new file mode 100644
index 0000000..f5120a2
--- /dev/null
+++ b/drivers/mailbox/mailbox-pdc/pdc_debug.h
@@ -0,0 +1,24 @@
+/*
+ * Copyright 2016 Broadcom
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, version 2, as
+ * published by the Free Software Foundation (the "GPL").
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License version 2 (GPLv2) for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * version 2 (GPLv2) along with this source code.
+ */
+
+#ifndef _PDC_DEBUG_H
+#define _PDC_DEBUG_H
+
+#include "pdc.h"
+
+void pdc_setup_debugfs(struct pdc_state *pdcs);
+void pdc_free_debugfs(void);
+#endif
diff --git a/include/linux/mailbox/brcm-message.h b/include/linux/mailbox/brcm-message.h
new file mode 100644
index 0000000..6b55c93
--- /dev/null
+++ b/include/linux/mailbox/brcm-message.h
@@ -0,0 +1,56 @@
+/*
+ *  Copyright (C) 2016 Broadcom
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * Common header for Broadcom mailbox messages which is shared across
+ * Broadcom SoCs and Broadcom mailbox client drivers.
+ */
+
+#ifndef _LINUX_BRCM_MESSAGE_H_
+#define _LINUX_BRCM_MESSAGE_H_
+
+#include <linux/scatterlist.h>
+
+enum brcm_message_type {
+	BRCM_MESSAGE_UNKNOWN = 0,
+	BRCM_MESSAGE_SPU,
+	BRCM_MESSAGE_SBA,
+	BRCM_MESSAGE_MAX,
+};
+
+struct brcm_sba_command {
+	u64 cmd;
+#define BRCM_SBA_CMD_TYPE_A		BIT(0)
+#define BRCM_SBA_CMD_TYPE_B		BIT(1)
+#define BRCM_SBA_CMD_TYPE_C		BIT(2)
+#define BRCM_SBA_CMD_HAS_RESP		BIT(3)
+#define BRCM_SBA_CMD_HAS_OUTPUT		BIT(4)
+	u64 flags;
+	dma_addr_t input;
+	size_t input_len;
+	dma_addr_t resp;
+	size_t resp_len;
+	dma_addr_t output;
+	size_t output_len;
+};
+
+struct brcm_message {
+	enum brcm_message_type type;
+	union {
+		struct {
+			struct scatterlist *src;
+			struct scatterlist *dst;
+		} spu;
+		struct {
+			struct brcm_sba_command *cmds;
+			unsigned int cmds_count;
+		} sba;
+	};
+	void *ctx;
+	int error;
+};
+
+#endif /* _LINUX_BRCM_MESSAGE_H_ */
-- 
2.1.0

Re: [PATCH 2/3] mailbox: Add Broadcom PDC mailbox driver

[Jassi Brar]

On Tue, May 24, 2016 at 11:37 PM, Rob Rice <rob.rice@broadcom.com> wrote:
> From: Rob Rice <rrice@broadcom.com>
>
> The Broadcom PDC mailbox driver is a mailbox controller that
> manages data transfers to and from one or more offload engines.
>
> Signed-off-by: Rob Rice <rob.rice@broadcom.com>
> Reviewed-by: Scott Branden <scott.branden@broadcom.com>
> Reviewed-by: Ray Jui <ray.jui@broadcom.com>
> ---
>  drivers/mailbox/Kconfig                 |    9 +
>  drivers/mailbox/Makefile                |    2 +
>  drivers/mailbox/mailbox-pdc/Makefile    |    8 +
>  drivers/mailbox/mailbox-pdc/pdc.c       | 1181 +++++++++++++++++++++++++++++++
>  drivers/mailbox/mailbox-pdc/pdc.h       |  303 ++++++++
>  drivers/mailbox/mailbox-pdc/pdc_debug.c |  101 +++
>  drivers/mailbox/mailbox-pdc/pdc_debug.h |   24 +
>
We already have mailbox api testing rig - mailbox-test, still if you
really need debugfs exposure please merge it in pdc.c especially when
pdc_debug.c is only 80lines and it always compiled and currently has
to export functions globally. Maybe fold everything in a single like
other platforms?

> diff --git a/drivers/mailbox/Kconfig b/drivers/mailbox/Kconfig
> index 5305923..19c8c9a 100644
> --- a/drivers/mailbox/Kconfig
> +++ b/drivers/mailbox/Kconfig
> @@ -123,4 +123,13 @@ config XGENE_SLIMPRO_MBOX
>           It is used to send short messages between ARM64-bit cores and
>           the SLIMpro Management Engine, primarily for PM. Say Y here if you
>           want to use the APM X-Gene SLIMpro IPCM support.
> +
> +config BCM_PDC_MBOX
> +       tristate "Broadcom PDC Mailbox"
> +       depends on ARM64
> +       default ARCH_BCM_IPROC
> +       help
> +         Mailbox implementation for the Broadcom PDC ring manager,
> +         which provides access to various offload engines on Broadcom
> +         SoCs. Say Y here if you want to use the Broadcom PDC.
>  endif
> diff --git a/drivers/mailbox/Makefile b/drivers/mailbox/Makefile
> index 0be3e74..2c14a03 100644
> --- a/drivers/mailbox/Makefile
> +++ b/drivers/mailbox/Makefile
> @@ -25,3 +25,5 @@ obj-$(CONFIG_TI_MESSAGE_MANAGER) += ti-msgmgr.o
>  obj-$(CONFIG_XGENE_SLIMPRO_MBOX) += mailbox-xgene-slimpro.o
>
>  obj-$(CONFIG_HI6220_MBOX)      += hi6220-mailbox.o
> +
> +obj-$(CONFIG_BCM_PDC_MBOX)     += mailbox-pdc/
> diff --git a/drivers/mailbox/mailbox-pdc/Makefile b/drivers/mailbox/mailbox-pdc/Makefile
> new file mode 100644
> index 0000000..5dc78c3
> --- /dev/null
> +++ b/drivers/mailbox/mailbox-pdc/Makefile
> @@ -0,0 +1,8 @@
> +# Makefile for Broadcom PDC Mailbox driver
> +
> +# Uncomment to enable debug tracing in the PDC driver
> +# CFLAGS_pdc.o := -DDEBUG
> +# CFLAGS_pdc_debug.o := -DDEBUG
> +
> +pdc-mbox-objs := pdc.o pdc_debug.o
> +obj-$(CONFIG_BCM_PDC_MBOX) := pdc-mbox.o
> diff --git a/drivers/mailbox/mailbox-pdc/pdc.c b/drivers/mailbox/mailbox-pdc/pdc.c
> new file mode 100644
> index 0000000..1c24ea8
> --- /dev/null
> +++ b/drivers/mailbox/mailbox-pdc/pdc.c
> @@ -0,0 +1,1181 @@
> +/*
> + * Copyright 2016 Broadcom
> + *
> + * This program is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License, version 2, as
> + * published by the Free Software Foundation (the "GPL").
> + *
> + * This program is distributed in the hope that it will be useful, but
> + * WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
> + * General Public License version 2 (GPLv2) for more details.
> + *
> + * You should have received a copy of the GNU General Public License
> + * version 2 (GPLv2) along with this source code.
> + */
> +
> +/*
> + * Broadcom PDC Mailbox Driver
> + * The PDC provides a ring based programming interface to one or more hardware
> + * offload engines. For example, the PDC driver works with both SPU-M and SPU2
> + * cryptographic offload hardware. In some chips the PDC is referred to as MDE.
> + *
> + * The PDC driver registers with the Linux mailbox framework as a mailbox
> + * controller, once for each PDC instance. Ring 0 for each PDC is registered as
> + * a mailbox channel. The PDC driver uses interrupts to determine when data
> + * transfers to and from an offload engine are complete. The PDC driver uses
> + * threaded IRQs so that response messages are handled outside of interrupt
> + * context.
> + *
> + * The PDC driver allows multiple messages to be pending in the descriptor
> + * rings. The tx_msg_start descriptor index indicates where the last message
> + * starts. The txin_numd value at this index indicates how many descriptor
> + * indexes make up the message. Similar state is kept on the receive side. When
> + * an rx interrupt indicates a response is ready, the PDC driver processes numd
> + * descriptors from the tx and rx ring, thus processing one response at a time.
> + */
> +
> +#include <linux/errno.h>
> +#include <linux/module.h>
> +#include <linux/init.h>
> +#include <linux/slab.h>
> +#include <linux/interrupt.h>
> +#include <linux/wait.h>
> +#include <linux/platform_device.h>
> +#include <linux/io.h>
> +#include <linux/of.h>
> +#include <linux/of_device.h>
> +#include <linux/of_address.h>
> +#include <linux/of_irq.h>
> +#include <linux/mailbox_controller.h>
> +#include <linux/mailbox/brcm-message.h>
> +#include <linux/scatterlist.h>
> +#include <linux/dma-direction.h>
> +#include <linux/dma-mapping.h>
> +#include <linux/dmapool.h>
> +
> +#include "pdc.h"
> +#include "pdc_debug.h"
> +
> +/* Length of BCM header at start of SPU msg, in bytes */
> +#define BCM_HDR_LEN  8
> +
> +/*
> + * PDC driver reserves ringset 0 on each SPU for its own use. The driver does
> + * not currently support use of multiple ringsets on a single PDC engine.
> + */
> +#define PDC_RINGSET  0
> +
> +/*
> + * Interrupt mask and status definitions. Enable interrupts for tx and rx on
> + * ring 0
> + */
> +#define PDC_XMTINT_0         (24 + PDC_RINGSET)
> +#define PDC_RCVINT_0         (16 + PDC_RINGSET)
> +#define PDC_XMTINTEN_0       BIT(PDC_XMTINT_0)
> +#define PDC_RCVINTEN_0       BIT(PDC_RCVINT_0)
> +#define PDC_INTMASK  (PDC_XMTINTEN_0 | PDC_RCVINTEN_0)
> +#define PDC_LAZY_FRAMECOUNT  1
> +#define PDC_LAZY_TIMEOUT     10000
> +#define PDC_LAZY_INT  (PDC_LAZY_TIMEOUT | (PDC_LAZY_FRAMECOUNT << 24))
> +#define PDC_INTMASK_OFFSET   0x24
> +#define PDC_INTSTATUS_OFFSET 0x20
> +#define PDC_RCVLAZY0_OFFSET  (0x30 + 4 * PDC_RINGSET)
> +
> +/*
> + * For SPU2, configure MDE_CKSUM_CONTROL to write 17 bytes of metadata
> + * before frame
> + */
> +#define PDC_SPU2_RESP_HDR_LEN  17
> +#define PDC_CKSUM_CTRL         BIT(27)
> +#define PDC_CKSUM_CTRL_OFFSET  0x400
> +
> +#define PDC_SPUM_RESP_HDR_LEN  32
> +
> +/*
> + * Sets the following bits for write to transmit control reg:
> + *  0    - XmtEn - enable activity on the tx channel
> + * 11    - PtyChkDisable - parity check is disabled
> + * 20:18 - BurstLen = 3 -> 2^7 = 128 byte data reads from memory
> + */
> +#define PDC_TX_CTL              0x000C0801
> +
> +/*
> + * Sets the following bits for write to receive control reg:
> + * 0     - RcvEn - enable activity on the rx channel
> + * 7:1   - RcvOffset - size in bytes of status region at start of rx frame buf
> + * 9     - SepRxHdrDescEn - place start of new frames only in descriptors
> + *                          that have StartOfFrame set
> + * 10    - OflowContinue - on rx FIFO overflow, clear rx fifo, discard all
> + *                         remaining bytes in current frame, report error
> + *                         in rx frame status for current frame
> + * 11    - PtyChkDisable - parity check is disabled
> + * 20:18 - BurstLen = 3 -> 2^7 = 128 byte data reads from memory
> + */
> +#define PDC_RX_CTL              0x000C0E01
> +
> +#define CRYPTO_D64_RS0_CD_MASK   ((PDC_RING_ENTRIES * RING_ENTRY_SIZE) - 1)
> +
> +/* descriptor flags */
> +#define D64_CTRL1_EOT   BIT(28)        /* end of descriptor table */
> +#define D64_CTRL1_IOC   BIT(29)        /* interrupt on complete */
> +#define D64_CTRL1_EOF   BIT(30)        /* end of frame */
> +#define D64_CTRL1_SOF   BIT(31)        /* start of frame */
> +
> +#define RX_STATUS_OVERFLOW       0x00800000
> +#define RX_STATUS_LEN            0x0000FFFF
> +
> +#define PDC_TXREGS_OFFSET  0x200
> +#define PDC_RXREGS_OFFSET  0x220
> +
> +/* Maximum size buffer the DMA engine can handle */
> +#define PDC_DMA_BUF_MAX 16384
> +
> +/* Global variables */
> +
> +struct pdc_globals {
> +       /* Actual number of SPUs in hardware, as reported by device tree */
> +       u32 num_spu;
> +};
> +
> +static struct pdc_globals pdcg;
> +
> +/**
> + * pdc_build_rxd() - Build DMA descriptor to receive SPU result.
> + * @pdcs:      PDC state for SPU that will generate result
> + * @dma_addr:  DMA address of buffer that descriptor is being built for
> + * @buf_len:   Length of the receive buffer, in bytes
> + * @flags:     Flags to be stored in descriptor
> + */
> +static inline void
> +pdc_build_rxd(struct pdc_state *pdcs, dma_addr_t dma_addr,
> +             u32 buf_len, u32 flags)
> +{
> +       struct device *dev = &pdcs->pdev->dev;
> +
> +       dev_dbg(dev,
> +               "Writing rx descriptor for PDC %u at index %u with length %u. flags %#x\n",
> +               pdcs->pdc_idx, pdcs->rxout, buf_len, flags);
> +
> +       iowrite32(lower_32_bits(dma_addr),
> +                 (void *)&pdcs->rxd_64[pdcs->rxout].addrlow);
> +       iowrite32(upper_32_bits(dma_addr),
> +                 (void *)&pdcs->rxd_64[pdcs->rxout].addrhigh);
> +       iowrite32(flags, (void *)&pdcs->rxd_64[pdcs->rxout].ctrl1);
> +       iowrite32(buf_len, (void *)&pdcs->rxd_64[pdcs->rxout].ctrl2);
> +       /* bump ring index and return */
> +       pdcs->rxout = NEXTRXD(pdcs->rxout, pdcs->nrxpost);
> +}
> +
> +/**
> + * pdc_build_txd() - Build a DMA descriptor to transmit a SPU request to
> + * hardware.
> + * @pdcs:        PDC state for the SPU that will process this request
> + * @dma_addr:    DMA address of packet to be transmitted
> + * @buf_len:     Length of tx buffer, in bytes
> + * @flags:       Flags to be stored in descriptor
> + */
> +static inline void
> +pdc_build_txd(struct pdc_state *pdcs, dma_addr_t dma_addr, u32 buf_len,
> +             u32 flags)
> +{
> +       struct device *dev = &pdcs->pdev->dev;
> +
> +       dev_dbg(dev,
> +               "Writing tx descriptor for PDC %u at index %u with length %u, flags %#x\n",
> +               pdcs->pdc_idx, pdcs->txout, buf_len, flags);
> +
> +       iowrite32(lower_32_bits(dma_addr),
> +                 (void *)&pdcs->txd_64[pdcs->txout].addrlow);
> +       iowrite32(upper_32_bits(dma_addr),
> +                 (void *)&pdcs->txd_64[pdcs->txout].addrhigh);
> +       iowrite32(flags, (void *)&pdcs->txd_64[pdcs->txout].ctrl1);
> +       iowrite32(buf_len, (void *)&pdcs->txd_64[pdcs->txout].ctrl2);
> +
> +       /* bump ring index and return */
> +       pdcs->txout = NEXTTXD(pdcs->txout, pdcs->ntxpost);
> +}
> +
> +/**
> + * pdc_receive() - Receive a response message from a given SPU.
> + * @pdcs:    PDC state for the SPU to receive from
> + * @mssg:    mailbox message to be returned to client
> + *
> + * When the return code indicates success, the response message is available in
> + * the receive buffers provided prior to submission of the request.
> + *
> + * Input:
> + *   pdcs - PDC state structure for the SPU to be polled
> + *   mssg - mailbox message to be returned to client. This function sets the
> + *         context pointer on the message to help the client associate the
> + *         response with a request.
> + *
> + * Return:  PDC_SUCCESS if one or more receive descriptors was processed
> + *          -EAGAIN indicates that no response message is available
> + *          -EIO an error occurred
> + */
> +static int
> +pdc_receive(struct pdc_state *pdcs, struct brcm_message *mssg)
> +{
> +       struct device *dev = &pdcs->pdev->dev;
> +       u32 len, rx_status;
> +       u32 num_frags;
> +       int i;
> +       u8 *resp_hdr;    /* virtual addr of start of resp message DMA header */
> +       u32 frags_rdy;   /* number of fragments ready to read */
> +       u32 rx_idx;      /* ring index of start of receive frame */
> +       dma_addr_t resp_hdr_daddr;
> +
> +       spin_lock(&pdcs->pdc_lock);
> +
> +       /*
> +        * return if a complete response message is not yet ready.
> +        * rxin_numd[rxin] is the number of fragments in the next msg
> +        * to read.
> +        */
> +       frags_rdy = NRXDACTIVE(pdcs->rxin, pdcs->last_rx_curr, pdcs->nrxpost);
> +       if ((frags_rdy == 0) || (frags_rdy < pdcs->rxin_numd[pdcs->rxin])) {
> +               /* See if the hw has written more fragments than we know */
> +               pdcs->last_rx_curr =
> +                   (ioread32((void *)&pdcs->rxregs_64->status0) &
> +                    CRYPTO_D64_RS0_CD_MASK) / RING_ENTRY_SIZE;
> +               frags_rdy = NRXDACTIVE(pdcs->rxin, pdcs->last_rx_curr,
> +                                      pdcs->nrxpost);
> +               if ((frags_rdy == 0) ||
> +                   (frags_rdy < pdcs->rxin_numd[pdcs->rxin])) {
> +                       /* No response ready */
> +                       spin_unlock(&pdcs->pdc_lock);
> +                       return -EAGAIN;
> +               }
> +               /* can't read descriptors/data until write index is read */
> +               rmb();
> +       }
> +
> +       num_frags = pdcs->txin_numd[pdcs->txin];
> +       dma_unmap_sg(dev, pdcs->src_sg[pdcs->txin],
> +                    sg_nents(pdcs->src_sg[pdcs->txin]), DMA_TO_DEVICE);
> +
> +       for (i = 0; i < num_frags; i++)
> +               pdcs->txin = NEXTTXD(pdcs->txin, pdcs->ntxpost);
> +
> +       dev_dbg(dev, "PDC %u reclaimed %d tx descriptors",
> +               pdcs->pdc_idx, num_frags);
> +
> +       rx_idx = pdcs->rxin;
> +       num_frags = pdcs->rxin_numd[rx_idx];
> +       /* Return opaque context with result */
> +       mssg->ctx = pdcs->rxp_ctx[rx_idx];
> +       pdcs->rxp_ctx[rx_idx] = NULL;
> +       resp_hdr = pdcs->resp_hdr[rx_idx];
> +       resp_hdr_daddr = pdcs->resp_hdr_daddr[rx_idx];
> +       dma_unmap_sg(dev, pdcs->dst_sg[rx_idx],
> +                    sg_nents(pdcs->dst_sg[rx_idx]), DMA_FROM_DEVICE);
> +
> +       for (i = 0; i < num_frags; i++)
> +               pdcs->rxin = NEXTRXD(pdcs->rxin, pdcs->nrxpost);
> +
> +       spin_unlock(&pdcs->pdc_lock);
> +
> +       dev_dbg(dev, "PDC %u reclaimed %d rx descriptors",
> +               pdcs->pdc_idx, num_frags);
> +
> +       dev_dbg(dev,
> +               "PDC %u txin %u, txout %u, rxin %u, rxout %u, last_rx_curr %u\n",
> +               pdcs->pdc_idx, pdcs->txin, pdcs->txout, pdcs->rxin,
> +               pdcs->rxout, pdcs->last_rx_curr);
> +
> +       if (pdcs->pdc_resp_hdr_len == PDC_SPUM_RESP_HDR_LEN) {
> +               /*
> +                * For SPU-M, get length of response msg and rx overflow status.
> +                */
> +               rx_status = *((u32 *)resp_hdr);
> +               len = rx_status & RX_STATUS_LEN;
> +               dev_dbg(dev,
> +                       "SPU response length %u bytes", len);
> +               if (unlikely(((rx_status & RX_STATUS_OVERFLOW) || (!len)))) {
> +                       if (rx_status & RX_STATUS_OVERFLOW) {
> +                               dev_err_ratelimited(dev,
> +                                                   "crypto receive overflow");
> +                               pdcs->rx_oflow++;
> +                       } else {
> +                               dev_info_ratelimited(dev, "crypto rx len = 0");
> +                       }
> +                       return -EIO;
> +               }
> +       }
> +
> +       dma_pool_free(pdcs->rx_buf_pool, resp_hdr, resp_hdr_daddr);
> +
> +       pdcs->pdc_replies++;
> +       /* if we read one or more rx descriptors, claim success */
> +       if (num_frags > 0)
> +               return PDC_SUCCESS;
> +       else
> +               return -EIO;
> +}
> +
> +/**
> + * pdc_tx_list_sg_add() - Add the buffers in a scatterlist to the transmit
> + * descriptors for a given SPU. The scatterlist buffers contain the data for a
> + * SPU request message.
> + * @spu_idx:   The index of the SPU to submit the request to, [0, max_spu)
> + * @sg:        Scatterlist whose buffers contain part of the SPU request
> + *
> + * If a scatterlist buffer is larger than PDC_DMA_BUF_MAX, multiple descriptors
> + * are written for that buffer, each <= PDC_DMA_BUF_MAX byte in length.
> + *
> + * Return: PDC_SUCCESS if successful
> + *         < 0 otherwise
> + */
> +static int pdc_tx_list_sg_add(struct pdc_state *pdcs, struct scatterlist *sg)
> +{
> +       u32 flags = 0;
> +       u32 eot;
> +       u32 tx_avail;
> +
> +       /*
> +        * Num descriptors needed. Conservatively assume we need a descriptor
> +        * for every entry in sg.
> +        */
> +       u32 num_desc;
> +       u32 desc_w = 0; /* Number of tx descriptors written */
> +       u32 bufcnt;     /* Number of bytes of buffer pointed to by descriptor */
> +       dma_addr_t databufptr;  /* DMA address to put in descriptor */
> +
> +       num_desc = (u32)sg_nents(sg);
> +
> +       /* check whether enough tx descriptors are available */
> +       tx_avail = pdcs->ntxpost - NTXDACTIVE(pdcs->txin, pdcs->txout,
> +                                             pdcs->ntxpost);
> +       if (unlikely(num_desc > tx_avail)) {
> +               pdcs->txnobuf++;
> +               return -ENOSPC;
> +       }
> +
> +       /* build tx descriptors */
> +       if (pdcs->tx_msg_start == pdcs->txout) {
> +               /* Start of frame */
> +               pdcs->txin_numd[pdcs->tx_msg_start] = 0;
> +               pdcs->src_sg[pdcs->txout] = sg;
> +               flags = D64_CTRL1_SOF;
> +       }
> +
> +       while (sg) {
> +               if (unlikely(pdcs->txout == (pdcs->ntxd - 1)))
> +                       eot = D64_CTRL1_EOT;
> +               else
> +                       eot = 0;
> +
> +               /*
> +                * If sg buffer larger than PDC limit, split across
> +                * multiple descriptors
> +                */
> +               bufcnt = sg_dma_len(sg);
> +               databufptr = sg_dma_address(sg);
> +               while (bufcnt > PDC_DMA_BUF_MAX) {
> +                       pdc_build_txd(pdcs, databufptr, PDC_DMA_BUF_MAX,
> +                                     flags | eot);
> +                       desc_w++;
> +                       bufcnt -= PDC_DMA_BUF_MAX;
> +                       databufptr += PDC_DMA_BUF_MAX;
> +                       if (unlikely(pdcs->txout == (pdcs->ntxd - 1)))
> +                               eot = D64_CTRL1_EOT;
> +                       else
> +                               eot = 0;
> +               }
> +               sg = sg_next(sg);
> +               if (!sg)
> +                       /* Writing last descriptor for frame */
> +                       flags |= (D64_CTRL1_EOF | D64_CTRL1_IOC);
> +               pdc_build_txd(pdcs, databufptr, bufcnt, flags | eot);
> +               desc_w++;
> +               /* Clear start of frame after first descriptor */
> +               flags &= ~D64_CTRL1_SOF;
> +       }
> +       pdcs->txin_numd[pdcs->tx_msg_start] += desc_w;
> +
> +       return PDC_SUCCESS;
> +}
> +
> +/**
> + * pdc_tx_list_final() - Initiate DMA transfer of last frame written to tx
> + * ring.
> + * @pdcs:  PDC state for SPU to process the request
> + *
> + * Sets the index of the last descriptor written in both the rx and tx ring.
> + *
> + * Return: PDC_SUCCESS
> + */
> +static int pdc_tx_list_final(struct pdc_state *pdcs)
> +{
> +       /*
> +        * write barrier to ensure all register writes are complete
> +        * before chip starts to process new request
> +        */
> +       wmb();
> +       iowrite32(pdcs->rxout << 4, (void *)&pdcs->rxregs_64->ptr);
> +       iowrite32(pdcs->txout << 4, (void *)&pdcs->txregs_64->ptr);
> +       pdcs->pdc_requests++;
> +
> +       return PDC_SUCCESS;
> +}
> +
> +/**
> + * pdc_rx_list_init() - Start a new receive descriptor list for a given PDC.
> + * @pdcs:   PDC state for SPU handling request
> + * @dst_sg: scatterlist providing rx buffers for response to be returned to
> + *         mailbox client
> + * @ctx:    Opaque context for this request
> + *
> + * Posts a single receive descriptor to hold the metadata that precedes a
> + * response. For example, with SPU-M, the metadata is a 32-byte DMA header and
> + * an 8-byte BCM header. Moves the msg_start descriptor indexes for both tx and
> + * rx to indicate the start of a new message.
> + *
> + * Return:  PDC_SUCCESS if successful
> + *          < 0 if an error (e.g., rx ring is full)
> + */
> +static int pdc_rx_list_init(struct pdc_state *pdcs, struct scatterlist *dst_sg,
> +                           void *ctx)
> +{
> +       u32 flags = 0;
> +       u32 rx_avail;
> +       u32 rx_pkt_cnt = 1;     /* Adding a single rx buffer */
> +       dma_addr_t daddr;
> +       void *vaddr;
> +
> +       rx_avail = pdcs->nrxpost - NRXDACTIVE(pdcs->rxin, pdcs->rxout,
> +                                             pdcs->nrxpost);
> +       if (unlikely(rx_pkt_cnt > rx_avail)) {
> +               pdcs->rxnobuf++;
> +               return -ENOSPC;
> +       }
> +
> +       /* allocate a buffer for the dma rx status */
> +       vaddr = dma_pool_zalloc(pdcs->rx_buf_pool, GFP_ATOMIC, &daddr);
> +       if (!vaddr)
> +               return -ENOMEM;
> +
> +       /*
> +        * Update msg_start indexes for both tx and rx to indicate the start
> +        * of a new sequence of descriptor indexes that contain the fragments
> +        * of the same message.
> +        */
> +       pdcs->rx_msg_start = pdcs->rxout;
> +       pdcs->tx_msg_start = pdcs->txout;
> +
> +       /* This is always the first descriptor in the receive sequence */
> +       flags = D64_CTRL1_SOF;
> +       pdcs->rxin_numd[pdcs->rx_msg_start] = 1;
> +
> +       if (unlikely(pdcs->rxout == (pdcs->nrxd - 1)))
> +               flags |= D64_CTRL1_EOT;
> +
> +       pdcs->rxp_ctx[pdcs->rxout] = ctx;
> +       pdcs->dst_sg[pdcs->rxout] = dst_sg;
> +       pdcs->resp_hdr[pdcs->rxout] = vaddr;
> +       pdcs->resp_hdr_daddr[pdcs->rxout] = daddr;
> +       pdc_build_rxd(pdcs, daddr, pdcs->pdc_resp_hdr_len, flags);
> +       return PDC_SUCCESS;
> +}
> +
> +/**
> + * pdc_rx_list_sg_add() - Add the buffers in a scatterlist to the receive
> + * descriptors for a given SPU. The caller must have already DMA mapped the
> + * scatterlist.
> + * @spu_idx:    Indicates which SPU the buffers are for
> + * @sg:         Scatterlist whose buffers are added to the receive ring
> + *
> + * If a receive buffer in the scatterlist is larger than PDC_DMA_BUF_MAX,
> + * multiple receive descriptors are written, each with a buffer <=
> + * PDC_DMA_BUF_MAX.
> + *
> + * Return: PDC_SUCCESS if successful
> + *         < 0 otherwise (e.g., receive ring is full)
> + */
> +static int pdc_rx_list_sg_add(struct pdc_state *pdcs, struct scatterlist *sg)
> +{
> +       u32 flags = 0;
> +       u32 rx_avail;
> +
> +       /*
> +        * Num descriptors needed. Conservatively assume we need a descriptor
> +        * for every entry from our starting point in the scatterlist.
> +        */
> +       u32 num_desc;
> +       u32 desc_w = 0; /* Number of tx descriptors written */
> +       u32 bufcnt;     /* Number of bytes of buffer pointed to by descriptor */
> +       dma_addr_t databufptr;  /* DMA address to put in descriptor */
> +
> +       num_desc = (u32)sg_nents(sg);
> +
> +       rx_avail = pdcs->nrxpost - NRXDACTIVE(pdcs->rxin, pdcs->rxout,
> +                                             pdcs->nrxpost);
> +       if (unlikely(num_desc > rx_avail)) {
> +               pdcs->rxnobuf++;
> +               return -ENOSPC;
> +       }
> +
> +       while (sg) {
> +               if (unlikely(pdcs->rxout == (pdcs->nrxd - 1)))
> +                       flags = D64_CTRL1_EOT;
> +               else
> +                       flags = 0;
> +
> +               /*
> +                * If sg buffer larger than PDC limit, split across
> +                * multiple descriptors
> +                */
> +               bufcnt = sg_dma_len(sg);
> +               databufptr = sg_dma_address(sg);
> +               while (bufcnt > PDC_DMA_BUF_MAX) {
> +                       pdc_build_rxd(pdcs, databufptr, PDC_DMA_BUF_MAX, flags);
> +                       desc_w++;
> +                       bufcnt -= PDC_DMA_BUF_MAX;
> +                       databufptr += PDC_DMA_BUF_MAX;
> +                       if (unlikely(pdcs->rxout == (pdcs->nrxd - 1)))
> +                               flags = D64_CTRL1_EOT;
> +                       else
> +                               flags = 0;
> +               }
> +               pdc_build_rxd(pdcs, databufptr, bufcnt, flags);
> +               desc_w++;
> +               sg = sg_next(sg);
> +       }
> +       pdcs->rxin_numd[pdcs->rx_msg_start] += desc_w;
> +
> +       return PDC_SUCCESS;
> +}
> +
> +/**
> + * pdc_irq_handler() - Interrupt handler called in interrupt context.
> + * @irq:      Interrupt number that has fired
> + * @cookie:   PDC state for DMA engine that generated the interrupt
> + *
> + * We have to clear the device interrupt status flags here. So cache the
> + * status for later use in the thread function. Other than that, just return
> + * WAKE_THREAD to invoke the thread function.
> + *
> + * Return: IRQ_WAKE_THREAD if interrupt is ours
> + *         IRQ_NONE otherwise
> + */
> +static irqreturn_t pdc_irq_handler(int irq, void *cookie)
> +{
> +       struct pdc_state *pdcs = cookie;
> +       u32 intstatus = ioread32(pdcs->pdc_reg_vbase + PDC_INTSTATUS_OFFSET);
> +
> +       if (intstatus & PDC_XMTINTEN_0)
> +               set_bit(PDC_XMTINT_0, &pdcs->intstatus);
> +       if (intstatus & PDC_RCVINTEN_0)
> +               set_bit(PDC_RCVINT_0, &pdcs->intstatus);
> +
> +       /* Clear interrupt flags in device */
> +       iowrite32(intstatus, pdcs->pdc_reg_vbase + PDC_INTSTATUS_OFFSET);
> +
> +       /* Wakeup IRQ thread */
> +       if (pdcs && (irq == pdcs->pdc_irq) && (intstatus & PDC_INTMASK))
> +               return IRQ_WAKE_THREAD;
> +
> +       return IRQ_NONE;
> +}
> +
> +/**
> + * pdc_irq_thread() - Function invoked on deferred thread when a DMA tx has
> + * completed or data is available to receive.
> + * @irq:    Interrupt number
> + * @cookie: PDC state for PDC that generated the interrupt
> + *
> + * On DMA tx complete, notify the mailbox client. On DMA rx complete, process
> + * as many SPU response messages as are available and send each to the mailbox
> + * client.
> + *
> + * Return: IRQ_HANDLED if we recognized and handled the interrupt
> + *         IRQ_NONE otherwise
> + */
> +static irqreturn_t pdc_irq_thread(int irq, void *cookie)
> +{
> +       struct pdc_state *pdcs = cookie;
> +       struct mbox_controller *mbc;
> +       struct mbox_chan *chan;
> +       bool tx_int;
> +       bool rx_int;
> +       int rx_status;
> +       struct brcm_message mssg;
> +
> +       tx_int = test_and_clear_bit(PDC_XMTINT_0, &pdcs->intstatus);
> +       rx_int = test_and_clear_bit(PDC_RCVINT_0, &pdcs->intstatus);
> +
> +       if (pdcs && (tx_int || rx_int)) {
> +               dev_dbg(&pdcs->pdev->dev,
> +                       "%s() got irq %d with tx_int %s, rx_int %s",
> +                       __func__, irq,
> +                       tx_int ? "set" : "clear", rx_int ? "set" : "clear");
> +
> +               mbc = &pdcs->mbc;
> +               chan = &mbc->chans[0];
> +
> +               if (tx_int) {
> +                       dev_dbg(&pdcs->pdev->dev, "%s(): tx done", __func__);
> +                       /* only one frame in flight at a time */
> +                       mbox_chan_txdone(chan, PDC_SUCCESS);
> +               }
> +               if (rx_int) {
> +                       while (1) {
> +                               /* Could be many frames ready */
> +                               memset(&mssg, 0, sizeof(mssg));
> +                               mssg.type = BRCM_MESSAGE_SPU;
> +                               rx_status = pdc_receive(pdcs, &mssg);
> +                               if (rx_status >= 0) {
> +                                       dev_dbg(&pdcs->pdev->dev,
> +                                               "%s(): invoking client rx cb",
> +                                               __func__);
> +                                       mbox_chan_received_data(chan, &mssg);
> +                               } else {
> +                                       dev_dbg(&pdcs->pdev->dev,
> +                                               "%s(): no SPU response available",
> +                                               __func__);
> +                                       break;
> +                               }
> +                       }
> +               }
> +               return IRQ_HANDLED;
> +       }
> +       return IRQ_NONE;
> +}
> +
> +/**
> + * pdc_ring_init() - Allocate DMA rings and initialize constant fields of
> + * descriptors in one ringset.
> + * @pdcs:    PDC instance state
> + * @ringset: index of ringset being used
> + *
> + * Return: PDC_SUCCESS if ring initialized
> + *         < 0 otherwise
> + */
> +static int pdc_ring_init(struct pdc_state *pdcs, int ringset)
> +{
> +       int i;
> +       int err = PDC_SUCCESS;
> +       struct dma64 *dma_reg;
> +       struct device *dev = &pdcs->pdev->dev;
> +       struct pdc_ring_alloc tx;
> +       struct pdc_ring_alloc rx;
> +
> +       /* Allocate tx ring */
> +       tx.vbase = dma_pool_zalloc(pdcs->ring_pool, GFP_KERNEL, &tx.dmabase);
> +       if (!tx.vbase) {
> +               err = -ENOMEM;
> +               goto done;
> +       }
> +
> +       /* Allocate rx ring */
> +       rx.vbase = dma_pool_zalloc(pdcs->ring_pool, GFP_KERNEL, &rx.dmabase);
> +       if (!rx.vbase) {
> +               err = -ENOMEM;
> +               goto fail_dealloc;
> +       }
> +
> +       dev_dbg(dev, " - base DMA addr of tx ring      %#llx", tx.dmabase);
> +       dev_dbg(dev, " - base virtual addr of tx ring  %p", tx.vbase);
> +       dev_dbg(dev, " - base DMA addr of rx ring      %#llx", rx.dmabase);
> +       dev_dbg(dev, " - base virtual addr of rx ring  %p", rx.vbase);
> +
> +       /* lock after ring allocation to avoid scheduling while atomic */
> +       spin_lock(&pdcs->pdc_lock);
> +
> +       memcpy(&pdcs->tx_ring_alloc, &tx, sizeof(tx));
> +       memcpy(&pdcs->rx_ring_alloc, &rx, sizeof(rx));
> +
> +       pdcs->rxin = 0;
> +       pdcs->rx_msg_start = 0;
> +       pdcs->last_rx_curr = 0;
> +       pdcs->rxout = 0;
> +       pdcs->txin = 0;
> +       pdcs->tx_msg_start = 0;
> +       pdcs->txout = 0;
> +
> +       /* Set descriptor array base addresses */
> +       pdcs->txd_64 = (struct dma64dd *)pdcs->tx_ring_alloc.vbase;
> +       pdcs->rxd_64 = (struct dma64dd *)pdcs->rx_ring_alloc.vbase;
> +
> +       /* Tell device the base DMA address of each ring */
> +       dma_reg = &pdcs->regs->dmaregs[ringset];
> +       iowrite32(lower_32_bits(pdcs->tx_ring_alloc.dmabase),
> +                 (void *)&dma_reg->dmaxmt.addrlow);
> +       iowrite32(upper_32_bits(pdcs->tx_ring_alloc.dmabase),
> +                 (void *)&dma_reg->dmaxmt.addrhigh);
> +
> +       iowrite32(lower_32_bits(pdcs->rx_ring_alloc.dmabase),
> +                 (void *)&dma_reg->dmarcv.addrlow);
> +       iowrite32(upper_32_bits(pdcs->rx_ring_alloc.dmabase),
> +                 (void *)&dma_reg->dmarcv.addrhigh);
> +
> +       /* Initialize descriptors */
> +       for (i = 0; i < PDC_RING_ENTRIES; i++) {
> +               /* Every tx descriptor can be used for start of frame. */
> +               if (i != pdcs->ntxpost) {
> +                       iowrite32(D64_CTRL1_SOF | D64_CTRL1_EOF,
> +                                 (void *)&pdcs->txd_64[i].ctrl1);
> +               } else {
> +                       /* Last descriptor in ringset. Set End of Table. */
> +                       iowrite32(D64_CTRL1_SOF | D64_CTRL1_EOF |
> +                                 D64_CTRL1_EOT,
> +                                 (void *)&pdcs->txd_64[i].ctrl1);
> +               }
> +
> +               /* Every rx descriptor can be used for start of frame */
> +               if (i != pdcs->nrxpost) {
> +                       iowrite32(D64_CTRL1_SOF,
> +                                 (void *)&pdcs->rxd_64[i].ctrl1);
> +               } else {
> +                       /* Last descriptor in ringset. Set End of Table. */
> +                       iowrite32(D64_CTRL1_SOF | D64_CTRL1_EOT,
> +                                 (void *)&pdcs->rxd_64[i].ctrl1);
> +               }
> +       }
> +       spin_unlock(&pdcs->pdc_lock);
> +       return PDC_SUCCESS;
> +
> +fail_dealloc:
> +       dma_pool_free(pdcs->ring_pool, tx.vbase, tx.dmabase);
> +done:
> +       return err;
> +}
> +
> +static void pdc_ring_free(struct pdc_state *pdcs)
> +{
> +       if (pdcs->tx_ring_alloc.vbase) {
> +               dma_pool_free(pdcs->ring_pool, pdcs->tx_ring_alloc.vbase,
> +                             pdcs->tx_ring_alloc.dmabase);
> +               pdcs->tx_ring_alloc.vbase = NULL;
> +       }
> +
> +       if (pdcs->rx_ring_alloc.vbase) {
> +               dma_pool_free(pdcs->ring_pool, pdcs->rx_ring_alloc.vbase,
> +                             pdcs->rx_ring_alloc.dmabase);
> +               pdcs->rx_ring_alloc.vbase = NULL;
> +       }
> +}
> +
> +/**
> + * pdc_send_data() - mailbox send_data function
> + * @chan:      The mailbox channel on which the data is sent. The channel
> + *              corresponds to a DMA ringset.
> + * @data:      The mailbox message to be sent. The message must be a
> + *              brcm_message structure.
> + *
> + * This function is registered as the send_data function for the mailbox
> + * controller. From the destination scatterlist in the mailbox message, it
> + * creates a sequence of receive descriptors in the rx ring. From the source
> + * scatterlist, it creates a sequence of transmit descriptors in the tx ring.
> + * After creating the descriptors, it writes the rx ptr and tx ptr registers to
> + * initiate the DMA transfer.
> + *
> + * This function does the DMA map and unmap of the src and dst scatterlists in
> + * the mailbox message.
> + *
> + * Return: 0 if successful
> + *        -ENOTSUPP if the mailbox message is a type this driver does not
> + *                     support
> + *         < 0 if an error
> + */
> +static int pdc_send_data(struct mbox_chan *chan, void *data)
> +{
> +       struct pdc_state *pdcs = chan->con_priv;
> +       struct device *dev = &pdcs->pdev->dev;
> +       struct brcm_message *mssg = data;
> +       int err = PDC_SUCCESS;
> +       int src_nent;
> +       int dst_nent;
> +       int nent;
> +
> +       if (mssg->type != BRCM_MESSAGE_SPU)
> +               return -ENOTSUPP;
> +
> +       src_nent = sg_nents(mssg->spu.src);
> +       if (src_nent) {
> +               nent = dma_map_sg(dev, mssg->spu.src, src_nent, DMA_TO_DEVICE);
> +               if (nent == 0)
> +                       return -EIO;
> +       }
> +
> +       dst_nent = sg_nents(mssg->spu.dst);
> +       if (dst_nent) {
> +               nent = dma_map_sg(dev, mssg->spu.dst, dst_nent,
> +                                 DMA_FROM_DEVICE);
> +               if (nent == 0) {
> +                       dma_unmap_sg(dev, mssg->spu.src, src_nent,
> +                                    DMA_TO_DEVICE);
> +                       return -EIO;
> +               }
> +       }
> +
> +       spin_lock(&pdcs->pdc_lock);
> +
> +       /* Create rx descriptors to SPU catch response */
> +       err = pdc_rx_list_init(pdcs, mssg->spu.dst, mssg->ctx);
> +       err |= pdc_rx_list_sg_add(pdcs, mssg->spu.dst);
> +
> +       /* Create tx descriptors to submit SPU request */
> +       err |= pdc_tx_list_sg_add(pdcs, mssg->spu.src);
> +       err |= pdc_tx_list_final(pdcs); /* initiate transfer */
> +
> +       spin_unlock(&pdcs->pdc_lock);
> +
> +       if (err)
> +               dev_err(&pdcs->pdev->dev,
> +                       "%s failed with error %d", __func__, err);
> +
> +       return err;
> +}
> +
> +static int pdc_startup(struct mbox_chan *chan)
> +{
> +       return pdc_ring_init(chan->con_priv, PDC_RINGSET);
> +}
> +
> +static void pdc_shutdown(struct mbox_chan *chan)
> +{
> +       struct pdc_state *pdcs = chan->con_priv;
> +
> +       if (pdcs)
> +               dev_dbg(&pdcs->pdev->dev,
> +                       "Shutdown mailbox channel for PDC %u", pdcs->pdc_idx);
> +
> +       pdc_ring_free(pdcs);
> +}
> +
> +/**
> + * pdc_hw_init() - Use the given initialization parameters to initialize the
> + * state for one of the PDCs.
> + * @dev:   device structure for PDC
> + * @pdcs:  state of the PDC
> + * @parms: parameter values to set
> + */
> +static
> +void pdc_hw_init(struct device *dev, struct pdc_state *pdcs,
> +                struct hw_init_parms *parms)
>
You could do without the dev and parms.

> +{
> +       struct dma64 *dma_reg;
> +
> +       int ringset = PDC_RINGSET;
> +
> +       dev_dbg(dev, "PDC %u initial values:", pdcs->pdc_idx);
> +       dev_dbg(dev, "state structure:                   %p",
> +               pdcs);
> +       dev_dbg(dev, " - base physical addr of hw regs   %#llx",
> +               parms->hw_pbase);
> +       dev_dbg(dev, " - base virtual addr of hw regs    %p",
> +               parms->hw_vbase);
> +
> +       /* initialize data structures */
> +       pdcs->regs = (struct pdc_regs *)parms->hw_vbase;
> +       pdcs->txregs_64 = (struct dma64_regs *)
> +           (void *)(((u8 *)parms->hw_vbase) +
> +                    PDC_TXREGS_OFFSET + (sizeof(struct dma64) * ringset));
> +       pdcs->rxregs_64 = (struct dma64_regs *)
> +           (void *)(((u8 *)parms->hw_vbase) +
> +                    PDC_RXREGS_OFFSET + (sizeof(struct dma64) * ringset));
> +
> +       pdcs->ntxd = parms->ring_entries;
> +       pdcs->nrxd = parms->ring_entries;
> +       pdcs->ntxpost = parms->ring_entries - 1;
> +       pdcs->nrxpost = parms->ring_entries - 1;
> +       pdcs->regs->intmask = 0;
> +
> +       dma_reg = &pdcs->regs->dmaregs[ringset];
> +       iowrite32(0, (void *)&dma_reg->dmaxmt.ptr);
> +       iowrite32(0, (void *)&dma_reg->dmarcv.ptr);
> +
> +       iowrite32(PDC_TX_CTL, (void *)&dma_reg->dmaxmt.control);
> +
> +       iowrite32(PDC_RX_CTL + (pdcs->rx_status_len << 1),
> +                 (void *)&dma_reg->dmarcv.control);
> +
> +       if (pdcs->pdc_resp_hdr_len == PDC_SPU2_RESP_HDR_LEN)
> +               iowrite32(PDC_CKSUM_CTRL,
> +                         pdcs->pdc_reg_vbase + PDC_CKSUM_CTRL_OFFSET);
> +}
> +
> +/**
> + * pdc_rx_buf_pool_create() - Pool of receive buffers used to catch the metadata
> + * header returned with each response message.
> + * @dev:  device structure
> + * @pdcs: PDC state structure
> + *
> + * The metadata is not returned to the mailbox client. So the PDC driver
> + * manages these buffers.
> + *
> + * Return: PDC_SUCCESS
> + *         -ENOMEM if pool creation fails
> + */
> +static int pdc_rx_buf_pool_create(struct device *dev, struct pdc_state *pdcs)
> +{
> +       pdcs->pdc_resp_hdr_len = pdcs->rx_status_len;
> +       if (pdcs->use_bcm_hdr)
> +               pdcs->pdc_resp_hdr_len += BCM_HDR_LEN;
> +
> +       pdcs->rx_buf_pool = dma_pool_create("pdc rx bufs", dev,
> +                                           pdcs->pdc_resp_hdr_len,
> +                                           RX_BUF_ALIGN, 0);
> +       if (!pdcs->rx_buf_pool)
> +               return -ENOMEM;
> +
> +       return PDC_SUCCESS;
> +}
> +
> +/**
> + * pdc_interrupts_init() - Initialize the interrupt configuration for a PDC and
> + * specify a threaded IRQ handler for deferred handling of interrupts outside of
> + * interrupt context.
> + * @pdev:   platform device
> + * @pdcs:   PDC state
> + *
> + * Set the interrupt mask for transmit and receive done.
> + * Set the lazy interrupt frame count to generate an interrupt for just one pkt.
> + *
> + * Return:  PDC_SUCCESS
> + *          <0 if threaded irq request fails
> + */
> +static int pdc_interrupts_init(struct platform_device *pdev,
> +                              struct pdc_state *pdcs)
> +{
> +       struct device *dev = &pdev->dev;
> +       struct device_node *dn = pdev->dev.of_node;
> +       int err;
> +
> +       pdcs->intstatus = 0;
> +
> +       /* interrupt configuration */
> +       iowrite32(PDC_INTMASK, pdcs->pdc_reg_vbase + PDC_INTMASK_OFFSET);
> +       iowrite32(PDC_LAZY_INT, pdcs->pdc_reg_vbase + PDC_RCVLAZY0_OFFSET);
> +
> +       /* read irq from device tree */
> +       pdcs->pdc_irq = irq_of_parse_and_map(dn, 0);
> +       dev_dbg(dev, "pdc device %s irq %u for pdcs %p",
> +               dev_name(dev), pdcs->pdc_irq, pdcs);
> +       err = devm_request_threaded_irq(dev, pdcs->pdc_irq,
> +                                       pdc_irq_handler,
> +                                       pdc_irq_thread, 0, dev_name(dev), pdcs);
> +       if (err) {
> +               dev_err(dev, "threaded tx IRQ %u request failed with err %d\n",
> +                       pdcs->pdc_irq, err);
> +               return err;
> +       }
> +       return PDC_SUCCESS;
> +}
> +
> +static const struct mbox_chan_ops pdc_mbox_chan_ops = {
> +       .send_data = pdc_send_data,
> +       .startup = pdc_startup,
> +       .shutdown = pdc_shutdown
> +};
> +
> +/**
> + * pdc_mb_init() - Initialize the mailbox controller.
> + * @pdcs:  PDC state
> + *
> + * Each PDC is a mailbox controller. Each ringset is a mailbox channel. Kernel
> + * driver only uses one ringset and thus one mb channel. PDC uses the transmit
> + * complete interrupt to determine when a mailbox message has successfully been
> + * transmitted.
> + *
> + * Return: 0 on success
> + *         < 0 if there is an allocation or registration failure
> + */
> +static int pdc_mb_init(struct pdc_state *pdcs)
> +{
> +       struct device *dev = &pdcs->pdev->dev;
> +       struct mbox_controller *mbc;
> +       int chan_index;
> +       int err;
> +
> +       mbc = &pdcs->mbc;
> +       mbc->dev = dev;
> +       mbc->ops = &pdc_mbox_chan_ops;
> +       mbc->num_chans = 1;
> +       mbc->chans = devm_kcalloc(dev, mbc->num_chans, sizeof(*mbc->chans),
> +                                 GFP_KERNEL);
> +       if (!mbc->chans)
> +               return -ENOMEM;
> +
> +       mbc->txdone_irq = true;
> +       mbc->txdone_poll = false;
> +       for (chan_index = 0; chan_index < mbc->num_chans; chan_index++)
> +               mbc->chans[chan_index].con_priv = pdcs;
> +
> +       /* Register mailbox controller */
> +       err = mbox_controller_register(mbc);
> +       if (err) {
> +               dev_crit(dev,
> +                        "Failed to register PDC mailbox controller. Error %d.",
> +                        err);
> +               return err;
> +       }
> +       return 0;
> +}
> +
> +/**
> + * pdc_dt_read() - Read application-specific data from device tree.
> + * @pdev:  Platform device
> + * @pdcs:  PDC state
> + *
> + * Reads the number of bytes of receive status that precede each received frame.
> + * Reads whether transmit and received frames should be preceded by an 8-byte
> + * BCM header.
> + *
> + * Return: 0 if successful
> + *         -ENODEV if device not available
> + */
> +static int pdc_dt_read(struct platform_device *pdev, struct pdc_state *pdcs)
> +{
> +       struct device *dev = &pdev->dev;
> +       struct device_node *dn = pdev->dev.of_node;
> +       int err;
> +
> +       err = of_property_read_u32(dn, "brcm,rx-status-len",
> +                                  &pdcs->rx_status_len);
> +       if (err < 0)
> +               dev_err(dev,
> +                       "%s failed to get DMA receive status length from device tree",
> +                       __func__);
> +
> +       pdcs->use_bcm_hdr = of_property_read_bool(dn, "brcm,use-bcm-hdr");
> +
> +       return 0;
> +}
> +
> +/**
> + * pdc_probe() - Probe function for PDC driver.
> + * @pdev:   PDC platform device
> + *
> + * Reserve and map register regions defined in device tree.
> + * Allocate and initialize tx and rx DMA rings.
> + * Initialize a mailbox controller for each PDC.
> + *
> + * Return: 0 if successful
> + *         < 0 if an error
> + */
> +static int pdc_probe(struct platform_device *pdev)
> +{
> +       int err = 0;
> +       struct device *dev = &pdev->dev;
> +       struct resource *pdc_regs;
> +       struct pdc_state *pdcs;
> +       struct hw_init_parms hw_parms;  /* params for initializing spu-dma */
> +
> +       /* PDC state for one SPU */
> +       pdcs = devm_kzalloc(dev, sizeof(*pdcs), GFP_KERNEL);
> +       if (!pdcs) {
> +               err = -ENOMEM;
> +               goto cleanup;
> +       }
> +
> +       spin_lock_init(&pdcs->pdc_lock);
> +       pdcs->pdev = pdev;
> +       platform_set_drvdata(pdev, pdcs);
> +       pdcs->pdc_idx = pdcg.num_spu;
> +       pdcg.num_spu++;
> +
> +       err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32));
> +       if (err) {
> +               dev_warn(dev, "PDC device cannot perform DMA. Error %d.", err);
> +               goto cleanup;
> +       }
> +
> +       /* Create DMA pool for tx ring */
> +       pdcs->ring_pool = dma_pool_create("pdc rings", dev, PDC_RING_SIZE,
> +                                         RING_ALIGN, 0);
> +       if (!pdcs->ring_pool) {
> +               err = -ENOMEM;
> +               goto cleanup;
> +       }
> +
> +       err = pdc_dt_read(pdev, pdcs);
> +       if (err)
> +               goto cleanup_ring_pool;
> +
> +       pdc_regs = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> +       if (!pdc_regs) {
> +               err = -ENODEV;
> +               goto cleanup_ring_pool;
> +       }
> +       dev_dbg(dev, "PDC register region res.start = %#llx, res.end = %#llx",
> +               pdc_regs->start, pdc_regs->end);
> +
> +       pdcs->pdc_reg_vbase = devm_ioremap_resource(&pdev->dev, pdc_regs);
> +       if (IS_ERR(pdcs->pdc_reg_vbase)) {
> +               err = PTR_ERR(pdcs->pdc_reg_vbase);
> +               dev_err(&pdev->dev, "Failed to map registers: %d\n", err);
> +               goto cleanup_ring_pool;
> +       }
> +
> +       /* create rx buffer pool after dt read to know how big buffers are */
> +       err = pdc_rx_buf_pool_create(dev, pdcs);
         'dev' argument could be omitted.

> +       if (err)
> +               goto cleanup_ring_pool;
> +
> +       hw_parms.hw_pbase = (dma_addr_t)(pdc_regs->start);
> +       hw_parms.hw_vbase = pdcs->pdc_reg_vbase;
> +
> +       hw_parms.ring_entries = PDC_RING_ENTRIES;
> +
> +       pdc_hw_init(dev, pdcs, &hw_parms);
     'dev' and 'hw_parms' argument could be omitted.

> +
> +       err = pdc_interrupts_init(pdev, pdcs);
     'pdev' argument could be omitted.


thanks.

Re: [PATCH 2/3] mailbox: Add Broadcom PDC mailbox driver

[Rob Rice]

Jassi,

Thanks for the comments. Replies below.

Rob

On Sun, Jun 26, 2016 at 2:31 AM, Jassi Brar <jassisinghbrar@gmail.com> wrote:
> On Tue, May 24, 2016 at 11:37 PM, Rob Rice <rob.rice@broadcom.com> wrote:
>> From: Rob Rice <rrice@broadcom.com>
>>
>> The Broadcom PDC mailbox driver is a mailbox controller that
>> manages data transfers to and from one or more offload engines.
>>
>> Signed-off-by: Rob Rice <rob.rice@broadcom.com>
>> Reviewed-by: Scott Branden <scott.branden@broadcom.com>
>> Reviewed-by: Ray Jui <ray.jui@broadcom.com>
>> ---
>>  drivers/mailbox/Kconfig                 |    9 +
>>  drivers/mailbox/Makefile                |    2 +
>>  drivers/mailbox/mailbox-pdc/Makefile    |    8 +
>>  drivers/mailbox/mailbox-pdc/pdc.c       | 1181 +++++++++++++++++++++++++++++++
>>  drivers/mailbox/mailbox-pdc/pdc.h       |  303 ++++++++
>>  drivers/mailbox/mailbox-pdc/pdc_debug.c |  101 +++
>>  drivers/mailbox/mailbox-pdc/pdc_debug.h |   24 +
>>
> We already have mailbox api testing rig - mailbox-test, still if you
> really need debugfs exposure please merge it in pdc.c especially when
> pdc_debug.c is only 80lines and it always compiled and currently has
> to export functions globally. Maybe fold everything in a single like
> other platforms?

[WRR] I'm using debugfs to display stats, not so much for testing the
module. So I don't think mailbox-test is a substitute. But if you
prefer, I'll move the stats code into pdc.c, along with the code from
pdc.h and move pdc.c from the mailbox-pdc directory up to
drivers/mailbox.

>
>> diff --git a/drivers/mailbox/Kconfig b/drivers/mailbox/Kconfig
>> index 5305923..19c8c9a 100644
>> --- a/drivers/mailbox/Kconfig
>> +++ b/drivers/mailbox/Kconfig
>> @@ -123,4 +123,13 @@ config XGENE_SLIMPRO_MBOX
>>           It is used to send short messages between ARM64-bit cores and
>>           the SLIMpro Management Engine, primarily for PM. Say Y here if you
>>           want to use the APM X-Gene SLIMpro IPCM support.
>> +
>> +config BCM_PDC_MBOX
>> +       tristate "Broadcom PDC Mailbox"
>> +       depends on ARM64
>> +       default ARCH_BCM_IPROC
>> +       help
>> +         Mailbox implementation for the Broadcom PDC ring manager,
>> +         which provides access to various offload engines on Broadcom
>> +         SoCs. Say Y here if you want to use the Broadcom PDC.
>>  endif
>> diff --git a/drivers/mailbox/Makefile b/drivers/mailbox/Makefile
>> index 0be3e74..2c14a03 100644
>> --- a/drivers/mailbox/Makefile
>> +++ b/drivers/mailbox/Makefile
>> @@ -25,3 +25,5 @@ obj-$(CONFIG_TI_MESSAGE_MANAGER) += ti-msgmgr.o
>>  obj-$(CONFIG_XGENE_SLIMPRO_MBOX) += mailbox-xgene-slimpro.o
>>
>>  obj-$(CONFIG_HI6220_MBOX)      += hi6220-mailbox.o
>> +
>> +obj-$(CONFIG_BCM_PDC_MBOX)     += mailbox-pdc/
>> diff --git a/drivers/mailbox/mailbox-pdc/Makefile b/drivers/mailbox/mailbox-pdc/Makefile
>> new file mode 100644
>> index 0000000..5dc78c3
>> --- /dev/null
>> +++ b/drivers/mailbox/mailbox-pdc/Makefile
>> @@ -0,0 +1,8 @@
>> +# Makefile for Broadcom PDC Mailbox driver
>> +
>> +# Uncomment to enable debug tracing in the PDC driver
>> +# CFLAGS_pdc.o := -DDEBUG
>> +# CFLAGS_pdc_debug.o := -DDEBUG
>> +
>> +pdc-mbox-objs := pdc.o pdc_debug.o
>> +obj-$(CONFIG_BCM_PDC_MBOX) := pdc-mbox.o
>> diff --git a/drivers/mailbox/mailbox-pdc/pdc.c b/drivers/mailbox/mailbox-pdc/pdc.c
>> new file mode 100644
>> index 0000000..1c24ea8
>> --- /dev/null
>> +++ b/drivers/mailbox/mailbox-pdc/pdc.c
>> @@ -0,0 +1,1181 @@
>> +/*
>> + * Copyright 2016 Broadcom
>> + *
>> + * This program is free software; you can redistribute it and/or modify
>> + * it under the terms of the GNU General Public License, version 2, as
>> + * published by the Free Software Foundation (the "GPL").
>> + *
>> + * This program is distributed in the hope that it will be useful, but
>> + * WITHOUT ANY WARRANTY; without even the implied warranty of
>> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
>> + * General Public License version 2 (GPLv2) for more details.
>> + *
>> + * You should have received a copy of the GNU General Public License
>> + * version 2 (GPLv2) along with this source code.
>> + */
>> +
>> +/*
>> + * Broadcom PDC Mailbox Driver
>> + * The PDC provides a ring based programming interface to one or more hardware
>> + * offload engines. For example, the PDC driver works with both SPU-M and SPU2
>> + * cryptographic offload hardware. In some chips the PDC is referred to as MDE.
>> + *
>> + * The PDC driver registers with the Linux mailbox framework as a mailbox
>> + * controller, once for each PDC instance. Ring 0 for each PDC is registered as
>> + * a mailbox channel. The PDC driver uses interrupts to determine when data
>> + * transfers to and from an offload engine are complete. The PDC driver uses
>> + * threaded IRQs so that response messages are handled outside of interrupt
>> + * context.
>> + *
>> + * The PDC driver allows multiple messages to be pending in the descriptor
>> + * rings. The tx_msg_start descriptor index indicates where the last message
>> + * starts. The txin_numd value at this index indicates how many descriptor
>> + * indexes make up the message. Similar state is kept on the receive side. When
>> + * an rx interrupt indicates a response is ready, the PDC driver processes numd
>> + * descriptors from the tx and rx ring, thus processing one response at a time.
>> + */
>> +
>> +#include <linux/errno.h>
>> +#include <linux/module.h>
>> +#include <linux/init.h>
>> +#include <linux/slab.h>
>> +#include <linux/interrupt.h>
>> +#include <linux/wait.h>
>> +#include <linux/platform_device.h>
>> +#include <linux/io.h>
>> +#include <linux/of.h>
>> +#include <linux/of_device.h>
>> +#include <linux/of_address.h>
>> +#include <linux/of_irq.h>
>> +#include <linux/mailbox_controller.h>
>> +#include <linux/mailbox/brcm-message.h>
>> +#include <linux/scatterlist.h>
>> +#include <linux/dma-direction.h>
>> +#include <linux/dma-mapping.h>
>> +#include <linux/dmapool.h>
>> +
>> +#include "pdc.h"
>> +#include "pdc_debug.h"
>> +
>> +/* Length of BCM header at start of SPU msg, in bytes */
>> +#define BCM_HDR_LEN  8
>> +
>> +/*
>> + * PDC driver reserves ringset 0 on each SPU for its own use. The driver does
>> + * not currently support use of multiple ringsets on a single PDC engine.
>> + */
>> +#define PDC_RINGSET  0
>> +
>> +/*
>> + * Interrupt mask and status definitions. Enable interrupts for tx and rx on
>> + * ring 0
>> + */
>> +#define PDC_XMTINT_0         (24 + PDC_RINGSET)
>> +#define PDC_RCVINT_0         (16 + PDC_RINGSET)
>> +#define PDC_XMTINTEN_0       BIT(PDC_XMTINT_0)
>> +#define PDC_RCVINTEN_0       BIT(PDC_RCVINT_0)
>> +#define PDC_INTMASK  (PDC_XMTINTEN_0 | PDC_RCVINTEN_0)
>> +#define PDC_LAZY_FRAMECOUNT  1
>> +#define PDC_LAZY_TIMEOUT     10000
>> +#define PDC_LAZY_INT  (PDC_LAZY_TIMEOUT | (PDC_LAZY_FRAMECOUNT << 24))
>> +#define PDC_INTMASK_OFFSET   0x24
>> +#define PDC_INTSTATUS_OFFSET 0x20
>> +#define PDC_RCVLAZY0_OFFSET  (0x30 + 4 * PDC_RINGSET)
>> +
>> +/*
>> + * For SPU2, configure MDE_CKSUM_CONTROL to write 17 bytes of metadata
>> + * before frame
>> + */
>> +#define PDC_SPU2_RESP_HDR_LEN  17
>> +#define PDC_CKSUM_CTRL         BIT(27)
>> +#define PDC_CKSUM_CTRL_OFFSET  0x400
>> +
>> +#define PDC_SPUM_RESP_HDR_LEN  32
>> +
>> +/*
>> + * Sets the following bits for write to transmit control reg:
>> + *  0    - XmtEn - enable activity on the tx channel
>> + * 11    - PtyChkDisable - parity check is disabled
>> + * 20:18 - BurstLen = 3 -> 2^7 = 128 byte data reads from memory
>> + */
>> +#define PDC_TX_CTL              0x000C0801
>> +
>> +/*
>> + * Sets the following bits for write to receive control reg:
>> + * 0     - RcvEn - enable activity on the rx channel
>> + * 7:1   - RcvOffset - size in bytes of status region at start of rx frame buf
>> + * 9     - SepRxHdrDescEn - place start of new frames only in descriptors
>> + *                          that have StartOfFrame set
>> + * 10    - OflowContinue - on rx FIFO overflow, clear rx fifo, discard all
>> + *                         remaining bytes in current frame, report error
>> + *                         in rx frame status for current frame
>> + * 11    - PtyChkDisable - parity check is disabled
>> + * 20:18 - BurstLen = 3 -> 2^7 = 128 byte data reads from memory
>> + */
>> +#define PDC_RX_CTL              0x000C0E01
>> +
>> +#define CRYPTO_D64_RS0_CD_MASK   ((PDC_RING_ENTRIES * RING_ENTRY_SIZE) - 1)
>> +
>> +/* descriptor flags */
>> +#define D64_CTRL1_EOT   BIT(28)        /* end of descriptor table */
>> +#define D64_CTRL1_IOC   BIT(29)        /* interrupt on complete */
>> +#define D64_CTRL1_EOF   BIT(30)        /* end of frame */
>> +#define D64_CTRL1_SOF   BIT(31)        /* start of frame */
>> +
>> +#define RX_STATUS_OVERFLOW       0x00800000
>> +#define RX_STATUS_LEN            0x0000FFFF
>> +
>> +#define PDC_TXREGS_OFFSET  0x200
>> +#define PDC_RXREGS_OFFSET  0x220
>> +
>> +/* Maximum size buffer the DMA engine can handle */
>> +#define PDC_DMA_BUF_MAX 16384
>> +
>> +/* Global variables */
>> +
>> +struct pdc_globals {
>> +       /* Actual number of SPUs in hardware, as reported by device tree */
>> +       u32 num_spu;
>> +};
>> +
>> +static struct pdc_globals pdcg;
>> +
>> +/**
>> + * pdc_build_rxd() - Build DMA descriptor to receive SPU result.
>> + * @pdcs:      PDC state for SPU that will generate result
>> + * @dma_addr:  DMA address of buffer that descriptor is being built for
>> + * @buf_len:   Length of the receive buffer, in bytes
>> + * @flags:     Flags to be stored in descriptor
>> + */
>> +static inline void
>> +pdc_build_rxd(struct pdc_state *pdcs, dma_addr_t dma_addr,
>> +             u32 buf_len, u32 flags)
>> +{
>> +       struct device *dev = &pdcs->pdev->dev;
>> +
>> +       dev_dbg(dev,
>> +               "Writing rx descriptor for PDC %u at index %u with length %u. flags %#x\n",
>> +               pdcs->pdc_idx, pdcs->rxout, buf_len, flags);
>> +
>> +       iowrite32(lower_32_bits(dma_addr),
>> +                 (void *)&pdcs->rxd_64[pdcs->rxout].addrlow);
>> +       iowrite32(upper_32_bits(dma_addr),
>> +                 (void *)&pdcs->rxd_64[pdcs->rxout].addrhigh);
>> +       iowrite32(flags, (void *)&pdcs->rxd_64[pdcs->rxout].ctrl1);
>> +       iowrite32(buf_len, (void *)&pdcs->rxd_64[pdcs->rxout].ctrl2);
>> +       /* bump ring index and return */
>> +       pdcs->rxout = NEXTRXD(pdcs->rxout, pdcs->nrxpost);
>> +}
>> +
>> +/**
>> + * pdc_build_txd() - Build a DMA descriptor to transmit a SPU request to
>> + * hardware.
>> + * @pdcs:        PDC state for the SPU that will process this request
>> + * @dma_addr:    DMA address of packet to be transmitted
>> + * @buf_len:     Length of tx buffer, in bytes
>> + * @flags:       Flags to be stored in descriptor
>> + */
>> +static inline void
>> +pdc_build_txd(struct pdc_state *pdcs, dma_addr_t dma_addr, u32 buf_len,
>> +             u32 flags)
>> +{
>> +       struct device *dev = &pdcs->pdev->dev;
>> +
>> +       dev_dbg(dev,
>> +               "Writing tx descriptor for PDC %u at index %u with length %u, flags %#x\n",
>> +               pdcs->pdc_idx, pdcs->txout, buf_len, flags);
>> +
>> +       iowrite32(lower_32_bits(dma_addr),
>> +                 (void *)&pdcs->txd_64[pdcs->txout].addrlow);
>> +       iowrite32(upper_32_bits(dma_addr),
>> +                 (void *)&pdcs->txd_64[pdcs->txout].addrhigh);
>> +       iowrite32(flags, (void *)&pdcs->txd_64[pdcs->txout].ctrl1);
>> +       iowrite32(buf_len, (void *)&pdcs->txd_64[pdcs->txout].ctrl2);
>> +
>> +       /* bump ring index and return */
>> +       pdcs->txout = NEXTTXD(pdcs->txout, pdcs->ntxpost);
>> +}
>> +
>> +/**
>> + * pdc_receive() - Receive a response message from a given SPU.
>> + * @pdcs:    PDC state for the SPU to receive from
>> + * @mssg:    mailbox message to be returned to client
>> + *
>> + * When the return code indicates success, the response message is available in
>> + * the receive buffers provided prior to submission of the request.
>> + *
>> + * Input:
>> + *   pdcs - PDC state structure for the SPU to be polled
>> + *   mssg - mailbox message to be returned to client. This function sets the
>> + *         context pointer on the message to help the client associate the
>> + *         response with a request.
>> + *
>> + * Return:  PDC_SUCCESS if one or more receive descriptors was processed
>> + *          -EAGAIN indicates that no response message is available
>> + *          -EIO an error occurred
>> + */
>> +static int
>> +pdc_receive(struct pdc_state *pdcs, struct brcm_message *mssg)
>> +{
>> +       struct device *dev = &pdcs->pdev->dev;
>> +       u32 len, rx_status;
>> +       u32 num_frags;
>> +       int i;
>> +       u8 *resp_hdr;    /* virtual addr of start of resp message DMA header */
>> +       u32 frags_rdy;   /* number of fragments ready to read */
>> +       u32 rx_idx;      /* ring index of start of receive frame */
>> +       dma_addr_t resp_hdr_daddr;
>> +
>> +       spin_lock(&pdcs->pdc_lock);
>> +
>> +       /*
>> +        * return if a complete response message is not yet ready.
>> +        * rxin_numd[rxin] is the number of fragments in the next msg
>> +        * to read.
>> +        */
>> +       frags_rdy = NRXDACTIVE(pdcs->rxin, pdcs->last_rx_curr, pdcs->nrxpost);
>> +       if ((frags_rdy == 0) || (frags_rdy < pdcs->rxin_numd[pdcs->rxin])) {
>> +               /* See if the hw has written more fragments than we know */
>> +               pdcs->last_rx_curr =
>> +                   (ioread32((void *)&pdcs->rxregs_64->status0) &
>> +                    CRYPTO_D64_RS0_CD_MASK) / RING_ENTRY_SIZE;
>> +               frags_rdy = NRXDACTIVE(pdcs->rxin, pdcs->last_rx_curr,
>> +                                      pdcs->nrxpost);
>> +               if ((frags_rdy == 0) ||
>> +                   (frags_rdy < pdcs->rxin_numd[pdcs->rxin])) {
>> +                       /* No response ready */
>> +                       spin_unlock(&pdcs->pdc_lock);
>> +                       return -EAGAIN;
>> +               }
>> +               /* can't read descriptors/data until write index is read */
>> +               rmb();
>> +       }
>> +
>> +       num_frags = pdcs->txin_numd[pdcs->txin];
>> +       dma_unmap_sg(dev, pdcs->src_sg[pdcs->txin],
>> +                    sg_nents(pdcs->src_sg[pdcs->txin]), DMA_TO_DEVICE);
>> +
>> +       for (i = 0; i < num_frags; i++)
>> +               pdcs->txin = NEXTTXD(pdcs->txin, pdcs->ntxpost);
>> +
>> +       dev_dbg(dev, "PDC %u reclaimed %d tx descriptors",
>> +               pdcs->pdc_idx, num_frags);
>> +
>> +       rx_idx = pdcs->rxin;
>> +       num_frags = pdcs->rxin_numd[rx_idx];
>> +       /* Return opaque context with result */
>> +       mssg->ctx = pdcs->rxp_ctx[rx_idx];
>> +       pdcs->rxp_ctx[rx_idx] = NULL;
>> +       resp_hdr = pdcs->resp_hdr[rx_idx];
>> +       resp_hdr_daddr = pdcs->resp_hdr_daddr[rx_idx];
>> +       dma_unmap_sg(dev, pdcs->dst_sg[rx_idx],
>> +                    sg_nents(pdcs->dst_sg[rx_idx]), DMA_FROM_DEVICE);
>> +
>> +       for (i = 0; i < num_frags; i++)
>> +               pdcs->rxin = NEXTRXD(pdcs->rxin, pdcs->nrxpost);
>> +
>> +       spin_unlock(&pdcs->pdc_lock);
>> +
>> +       dev_dbg(dev, "PDC %u reclaimed %d rx descriptors",
>> +               pdcs->pdc_idx, num_frags);
>> +
>> +       dev_dbg(dev,
>> +               "PDC %u txin %u, txout %u, rxin %u, rxout %u, last_rx_curr %u\n",
>> +               pdcs->pdc_idx, pdcs->txin, pdcs->txout, pdcs->rxin,
>> +               pdcs->rxout, pdcs->last_rx_curr);
>> +
>> +       if (pdcs->pdc_resp_hdr_len == PDC_SPUM_RESP_HDR_LEN) {
>> +               /*
>> +                * For SPU-M, get length of response msg and rx overflow status.
>> +                */
>> +               rx_status = *((u32 *)resp_hdr);
>> +               len = rx_status & RX_STATUS_LEN;
>> +               dev_dbg(dev,
>> +                       "SPU response length %u bytes", len);
>> +               if (unlikely(((rx_status & RX_STATUS_OVERFLOW) || (!len)))) {
>> +                       if (rx_status & RX_STATUS_OVERFLOW) {
>> +                               dev_err_ratelimited(dev,
>> +                                                   "crypto receive overflow");
>> +                               pdcs->rx_oflow++;
>> +                       } else {
>> +                               dev_info_ratelimited(dev, "crypto rx len = 0");
>> +                       }
>> +                       return -EIO;
>> +               }
>> +       }
>> +
>> +       dma_pool_free(pdcs->rx_buf_pool, resp_hdr, resp_hdr_daddr);
>> +
>> +       pdcs->pdc_replies++;
>> +       /* if we read one or more rx descriptors, claim success */
>> +       if (num_frags > 0)
>> +               return PDC_SUCCESS;
>> +       else
>> +               return -EIO;
>> +}
>> +
>> +/**
>> + * pdc_tx_list_sg_add() - Add the buffers in a scatterlist to the transmit
>> + * descriptors for a given SPU. The scatterlist buffers contain the data for a
>> + * SPU request message.
>> + * @spu_idx:   The index of the SPU to submit the request to, [0, max_spu)
>> + * @sg:        Scatterlist whose buffers contain part of the SPU request
>> + *
>> + * If a scatterlist buffer is larger than PDC_DMA_BUF_MAX, multiple descriptors
>> + * are written for that buffer, each <= PDC_DMA_BUF_MAX byte in length.
>> + *
>> + * Return: PDC_SUCCESS if successful
>> + *         < 0 otherwise
>> + */
>> +static int pdc_tx_list_sg_add(struct pdc_state *pdcs, struct scatterlist *sg)
>> +{
>> +       u32 flags = 0;
>> +       u32 eot;
>> +       u32 tx_avail;
>> +
>> +       /*
>> +        * Num descriptors needed. Conservatively assume we need a descriptor
>> +        * for every entry in sg.
>> +        */
>> +       u32 num_desc;
>> +       u32 desc_w = 0; /* Number of tx descriptors written */
>> +       u32 bufcnt;     /* Number of bytes of buffer pointed to by descriptor */
>> +       dma_addr_t databufptr;  /* DMA address to put in descriptor */
>> +
>> +       num_desc = (u32)sg_nents(sg);
>> +
>> +       /* check whether enough tx descriptors are available */
>> +       tx_avail = pdcs->ntxpost - NTXDACTIVE(pdcs->txin, pdcs->txout,
>> +                                             pdcs->ntxpost);
>> +       if (unlikely(num_desc > tx_avail)) {
>> +               pdcs->txnobuf++;
>> +               return -ENOSPC;
>> +       }
>> +
>> +       /* build tx descriptors */
>> +       if (pdcs->tx_msg_start == pdcs->txout) {
>> +               /* Start of frame */
>> +               pdcs->txin_numd[pdcs->tx_msg_start] = 0;
>> +               pdcs->src_sg[pdcs->txout] = sg;
>> +               flags = D64_CTRL1_SOF;
>> +       }
>> +
>> +       while (sg) {
>> +               if (unlikely(pdcs->txout == (pdcs->ntxd - 1)))
>> +                       eot = D64_CTRL1_EOT;
>> +               else
>> +                       eot = 0;
>> +
>> +               /*
>> +                * If sg buffer larger than PDC limit, split across
>> +                * multiple descriptors
>> +                */
>> +               bufcnt = sg_dma_len(sg);
>> +               databufptr = sg_dma_address(sg);
>> +               while (bufcnt > PDC_DMA_BUF_MAX) {
>> +                       pdc_build_txd(pdcs, databufptr, PDC_DMA_BUF_MAX,
>> +                                     flags | eot);
>> +                       desc_w++;
>> +                       bufcnt -= PDC_DMA_BUF_MAX;
>> +                       databufptr += PDC_DMA_BUF_MAX;
>> +                       if (unlikely(pdcs->txout == (pdcs->ntxd - 1)))
>> +                               eot = D64_CTRL1_EOT;
>> +                       else
>> +                               eot = 0;
>> +               }
>> +               sg = sg_next(sg);
>> +               if (!sg)
>> +                       /* Writing last descriptor for frame */
>> +                       flags |= (D64_CTRL1_EOF | D64_CTRL1_IOC);
>> +               pdc_build_txd(pdcs, databufptr, bufcnt, flags | eot);
>> +               desc_w++;
>> +               /* Clear start of frame after first descriptor */
>> +               flags &= ~D64_CTRL1_SOF;
>> +       }
>> +       pdcs->txin_numd[pdcs->tx_msg_start] += desc_w;
>> +
>> +       return PDC_SUCCESS;
>> +}
>> +
>> +/**
>> + * pdc_tx_list_final() - Initiate DMA transfer of last frame written to tx
>> + * ring.
>> + * @pdcs:  PDC state for SPU to process the request
>> + *
>> + * Sets the index of the last descriptor written in both the rx and tx ring.
>> + *
>> + * Return: PDC_SUCCESS
>> + */
>> +static int pdc_tx_list_final(struct pdc_state *pdcs)
>> +{
>> +       /*
>> +        * write barrier to ensure all register writes are complete
>> +        * before chip starts to process new request
>> +        */
>> +       wmb();
>> +       iowrite32(pdcs->rxout << 4, (void *)&pdcs->rxregs_64->ptr);
>> +       iowrite32(pdcs->txout << 4, (void *)&pdcs->txregs_64->ptr);
>> +       pdcs->pdc_requests++;
>> +
>> +       return PDC_SUCCESS;
>> +}
>> +
>> +/**
>> + * pdc_rx_list_init() - Start a new receive descriptor list for a given PDC.
>> + * @pdcs:   PDC state for SPU handling request
>> + * @dst_sg: scatterlist providing rx buffers for response to be returned to
>> + *         mailbox client
>> + * @ctx:    Opaque context for this request
>> + *
>> + * Posts a single receive descriptor to hold the metadata that precedes a
>> + * response. For example, with SPU-M, the metadata is a 32-byte DMA header and
>> + * an 8-byte BCM header. Moves the msg_start descriptor indexes for both tx and
>> + * rx to indicate the start of a new message.
>> + *
>> + * Return:  PDC_SUCCESS if successful
>> + *          < 0 if an error (e.g., rx ring is full)
>> + */
>> +static int pdc_rx_list_init(struct pdc_state *pdcs, struct scatterlist *dst_sg,
>> +                           void *ctx)
>> +{
>> +       u32 flags = 0;
>> +       u32 rx_avail;
>> +       u32 rx_pkt_cnt = 1;     /* Adding a single rx buffer */
>> +       dma_addr_t daddr;
>> +       void *vaddr;
>> +
>> +       rx_avail = pdcs->nrxpost - NRXDACTIVE(pdcs->rxin, pdcs->rxout,
>> +                                             pdcs->nrxpost);
>> +       if (unlikely(rx_pkt_cnt > rx_avail)) {
>> +               pdcs->rxnobuf++;
>> +               return -ENOSPC;
>> +       }
>> +
>> +       /* allocate a buffer for the dma rx status */
>> +       vaddr = dma_pool_zalloc(pdcs->rx_buf_pool, GFP_ATOMIC, &daddr);
>> +       if (!vaddr)
>> +               return -ENOMEM;
>> +
>> +       /*
>> +        * Update msg_start indexes for both tx and rx to indicate the start
>> +        * of a new sequence of descriptor indexes that contain the fragments
>> +        * of the same message.
>> +        */
>> +       pdcs->rx_msg_start = pdcs->rxout;
>> +       pdcs->tx_msg_start = pdcs->txout;
>> +
>> +       /* This is always the first descriptor in the receive sequence */
>> +       flags = D64_CTRL1_SOF;
>> +       pdcs->rxin_numd[pdcs->rx_msg_start] = 1;
>> +
>> +       if (unlikely(pdcs->rxout == (pdcs->nrxd - 1)))
>> +               flags |= D64_CTRL1_EOT;
>> +
>> +       pdcs->rxp_ctx[pdcs->rxout] = ctx;
>> +       pdcs->dst_sg[pdcs->rxout] = dst_sg;
>> +       pdcs->resp_hdr[pdcs->rxout] = vaddr;
>> +       pdcs->resp_hdr_daddr[pdcs->rxout] = daddr;
>> +       pdc_build_rxd(pdcs, daddr, pdcs->pdc_resp_hdr_len, flags);
>> +       return PDC_SUCCESS;
>> +}
>> +
>> +/**
>> + * pdc_rx_list_sg_add() - Add the buffers in a scatterlist to the receive
>> + * descriptors for a given SPU. The caller must have already DMA mapped the
>> + * scatterlist.
>> + * @spu_idx:    Indicates which SPU the buffers are for
>> + * @sg:         Scatterlist whose buffers are added to the receive ring
>> + *
>> + * If a receive buffer in the scatterlist is larger than PDC_DMA_BUF_MAX,
>> + * multiple receive descriptors are written, each with a buffer <=
>> + * PDC_DMA_BUF_MAX.
>> + *
>> + * Return: PDC_SUCCESS if successful
>> + *         < 0 otherwise (e.g., receive ring is full)
>> + */
>> +static int pdc_rx_list_sg_add(struct pdc_state *pdcs, struct scatterlist *sg)
>> +{
>> +       u32 flags = 0;
>> +       u32 rx_avail;
>> +
>> +       /*
>> +        * Num descriptors needed. Conservatively assume we need a descriptor
>> +        * for every entry from our starting point in the scatterlist.
>> +        */
>> +       u32 num_desc;
>> +       u32 desc_w = 0; /* Number of tx descriptors written */
>> +       u32 bufcnt;     /* Number of bytes of buffer pointed to by descriptor */
>> +       dma_addr_t databufptr;  /* DMA address to put in descriptor */
>> +
>> +       num_desc = (u32)sg_nents(sg);
>> +
>> +       rx_avail = pdcs->nrxpost - NRXDACTIVE(pdcs->rxin, pdcs->rxout,
>> +                                             pdcs->nrxpost);
>> +       if (unlikely(num_desc > rx_avail)) {
>> +               pdcs->rxnobuf++;
>> +               return -ENOSPC;
>> +       }
>> +
>> +       while (sg) {
>> +               if (unlikely(pdcs->rxout == (pdcs->nrxd - 1)))
>> +                       flags = D64_CTRL1_EOT;
>> +               else
>> +                       flags = 0;
>> +
>> +               /*
>> +                * If sg buffer larger than PDC limit, split across
>> +                * multiple descriptors
>> +                */
>> +               bufcnt = sg_dma_len(sg);
>> +               databufptr = sg_dma_address(sg);
>> +               while (bufcnt > PDC_DMA_BUF_MAX) {
>> +                       pdc_build_rxd(pdcs, databufptr, PDC_DMA_BUF_MAX, flags);
>> +                       desc_w++;
>> +                       bufcnt -= PDC_DMA_BUF_MAX;
>> +                       databufptr += PDC_DMA_BUF_MAX;
>> +                       if (unlikely(pdcs->rxout == (pdcs->nrxd - 1)))
>> +                               flags = D64_CTRL1_EOT;
>> +                       else
>> +                               flags = 0;
>> +               }
>> +               pdc_build_rxd(pdcs, databufptr, bufcnt, flags);
>> +               desc_w++;
>> +               sg = sg_next(sg);
>> +       }
>> +       pdcs->rxin_numd[pdcs->rx_msg_start] += desc_w;
>> +
>> +       return PDC_SUCCESS;
>> +}
>> +
>> +/**
>> + * pdc_irq_handler() - Interrupt handler called in interrupt context.
>> + * @irq:      Interrupt number that has fired
>> + * @cookie:   PDC state for DMA engine that generated the interrupt
>> + *
>> + * We have to clear the device interrupt status flags here. So cache the
>> + * status for later use in the thread function. Other than that, just return
>> + * WAKE_THREAD to invoke the thread function.
>> + *
>> + * Return: IRQ_WAKE_THREAD if interrupt is ours
>> + *         IRQ_NONE otherwise
>> + */
>> +static irqreturn_t pdc_irq_handler(int irq, void *cookie)
>> +{
>> +       struct pdc_state *pdcs = cookie;
>> +       u32 intstatus = ioread32(pdcs->pdc_reg_vbase + PDC_INTSTATUS_OFFSET);
>> +
>> +       if (intstatus & PDC_XMTINTEN_0)
>> +               set_bit(PDC_XMTINT_0, &pdcs->intstatus);
>> +       if (intstatus & PDC_RCVINTEN_0)
>> +               set_bit(PDC_RCVINT_0, &pdcs->intstatus);
>> +
>> +       /* Clear interrupt flags in device */
>> +       iowrite32(intstatus, pdcs->pdc_reg_vbase + PDC_INTSTATUS_OFFSET);
>> +
>> +       /* Wakeup IRQ thread */
>> +       if (pdcs && (irq == pdcs->pdc_irq) && (intstatus & PDC_INTMASK))
>> +               return IRQ_WAKE_THREAD;
>> +
>> +       return IRQ_NONE;
>> +}
>> +
>> +/**
>> + * pdc_irq_thread() - Function invoked on deferred thread when a DMA tx has
>> + * completed or data is available to receive.
>> + * @irq:    Interrupt number
>> + * @cookie: PDC state for PDC that generated the interrupt
>> + *
>> + * On DMA tx complete, notify the mailbox client. On DMA rx complete, process
>> + * as many SPU response messages as are available and send each to the mailbox
>> + * client.
>> + *
>> + * Return: IRQ_HANDLED if we recognized and handled the interrupt
>> + *         IRQ_NONE otherwise
>> + */
>> +static irqreturn_t pdc_irq_thread(int irq, void *cookie)
>> +{
>> +       struct pdc_state *pdcs = cookie;
>> +       struct mbox_controller *mbc;
>> +       struct mbox_chan *chan;
>> +       bool tx_int;
>> +       bool rx_int;
>> +       int rx_status;
>> +       struct brcm_message mssg;
>> +
>> +       tx_int = test_and_clear_bit(PDC_XMTINT_0, &pdcs->intstatus);
>> +       rx_int = test_and_clear_bit(PDC_RCVINT_0, &pdcs->intstatus);
>> +
>> +       if (pdcs && (tx_int || rx_int)) {
>> +               dev_dbg(&pdcs->pdev->dev,
>> +                       "%s() got irq %d with tx_int %s, rx_int %s",
>> +                       __func__, irq,
>> +                       tx_int ? "set" : "clear", rx_int ? "set" : "clear");
>> +
>> +               mbc = &pdcs->mbc;
>> +               chan = &mbc->chans[0];
>> +
>> +               if (tx_int) {
>> +                       dev_dbg(&pdcs->pdev->dev, "%s(): tx done", __func__);
>> +                       /* only one frame in flight at a time */
>> +                       mbox_chan_txdone(chan, PDC_SUCCESS);
>> +               }
>> +               if (rx_int) {
>> +                       while (1) {
>> +                               /* Could be many frames ready */
>> +                               memset(&mssg, 0, sizeof(mssg));
>> +                               mssg.type = BRCM_MESSAGE_SPU;
>> +                               rx_status = pdc_receive(pdcs, &mssg);
>> +                               if (rx_status >= 0) {
>> +                                       dev_dbg(&pdcs->pdev->dev,
>> +                                               "%s(): invoking client rx cb",
>> +                                               __func__);
>> +                                       mbox_chan_received_data(chan, &mssg);
>> +                               } else {
>> +                                       dev_dbg(&pdcs->pdev->dev,
>> +                                               "%s(): no SPU response available",
>> +                                               __func__);
>> +                                       break;
>> +                               }
>> +                       }
>> +               }
>> +               return IRQ_HANDLED;
>> +       }
>> +       return IRQ_NONE;
>> +}
>> +
>> +/**
>> + * pdc_ring_init() - Allocate DMA rings and initialize constant fields of
>> + * descriptors in one ringset.
>> + * @pdcs:    PDC instance state
>> + * @ringset: index of ringset being used
>> + *
>> + * Return: PDC_SUCCESS if ring initialized
>> + *         < 0 otherwise
>> + */
>> +static int pdc_ring_init(struct pdc_state *pdcs, int ringset)
>> +{
>> +       int i;
>> +       int err = PDC_SUCCESS;
>> +       struct dma64 *dma_reg;
>> +       struct device *dev = &pdcs->pdev->dev;
>> +       struct pdc_ring_alloc tx;
>> +       struct pdc_ring_alloc rx;
>> +
>> +       /* Allocate tx ring */
>> +       tx.vbase = dma_pool_zalloc(pdcs->ring_pool, GFP_KERNEL, &tx.dmabase);
>> +       if (!tx.vbase) {
>> +               err = -ENOMEM;
>> +               goto done;
>> +       }
>> +
>> +       /* Allocate rx ring */
>> +       rx.vbase = dma_pool_zalloc(pdcs->ring_pool, GFP_KERNEL, &rx.dmabase);
>> +       if (!rx.vbase) {
>> +               err = -ENOMEM;
>> +               goto fail_dealloc;
>> +       }
>> +
>> +       dev_dbg(dev, " - base DMA addr of tx ring      %#llx", tx.dmabase);
>> +       dev_dbg(dev, " - base virtual addr of tx ring  %p", tx.vbase);
>> +       dev_dbg(dev, " - base DMA addr of rx ring      %#llx", rx.dmabase);
>> +       dev_dbg(dev, " - base virtual addr of rx ring  %p", rx.vbase);
>> +
>> +       /* lock after ring allocation to avoid scheduling while atomic */
>> +       spin_lock(&pdcs->pdc_lock);
>> +
>> +       memcpy(&pdcs->tx_ring_alloc, &tx, sizeof(tx));
>> +       memcpy(&pdcs->rx_ring_alloc, &rx, sizeof(rx));
>> +
>> +       pdcs->rxin = 0;
>> +       pdcs->rx_msg_start = 0;
>> +       pdcs->last_rx_curr = 0;
>> +       pdcs->rxout = 0;
>> +       pdcs->txin = 0;
>> +       pdcs->tx_msg_start = 0;
>> +       pdcs->txout = 0;
>> +
>> +       /* Set descriptor array base addresses */
>> +       pdcs->txd_64 = (struct dma64dd *)pdcs->tx_ring_alloc.vbase;
>> +       pdcs->rxd_64 = (struct dma64dd *)pdcs->rx_ring_alloc.vbase;
>> +
>> +       /* Tell device the base DMA address of each ring */
>> +       dma_reg = &pdcs->regs->dmaregs[ringset];
>> +       iowrite32(lower_32_bits(pdcs->tx_ring_alloc.dmabase),
>> +                 (void *)&dma_reg->dmaxmt.addrlow);
>> +       iowrite32(upper_32_bits(pdcs->tx_ring_alloc.dmabase),
>> +                 (void *)&dma_reg->dmaxmt.addrhigh);
>> +
>> +       iowrite32(lower_32_bits(pdcs->rx_ring_alloc.dmabase),
>> +                 (void *)&dma_reg->dmarcv.addrlow);
>> +       iowrite32(upper_32_bits(pdcs->rx_ring_alloc.dmabase),
>> +                 (void *)&dma_reg->dmarcv.addrhigh);
>> +
>> +       /* Initialize descriptors */
>> +       for (i = 0; i < PDC_RING_ENTRIES; i++) {
>> +               /* Every tx descriptor can be used for start of frame. */
>> +               if (i != pdcs->ntxpost) {
>> +                       iowrite32(D64_CTRL1_SOF | D64_CTRL1_EOF,
>> +                                 (void *)&pdcs->txd_64[i].ctrl1);
>> +               } else {
>> +                       /* Last descriptor in ringset. Set End of Table. */
>> +                       iowrite32(D64_CTRL1_SOF | D64_CTRL1_EOF |
>> +                                 D64_CTRL1_EOT,
>> +                                 (void *)&pdcs->txd_64[i].ctrl1);
>> +               }
>> +
>> +               /* Every rx descriptor can be used for start of frame */
>> +               if (i != pdcs->nrxpost) {
>> +                       iowrite32(D64_CTRL1_SOF,
>> +                                 (void *)&pdcs->rxd_64[i].ctrl1);
>> +               } else {
>> +                       /* Last descriptor in ringset. Set End of Table. */
>> +                       iowrite32(D64_CTRL1_SOF | D64_CTRL1_EOT,
>> +                                 (void *)&pdcs->rxd_64[i].ctrl1);
>> +               }
>> +       }
>> +       spin_unlock(&pdcs->pdc_lock);
>> +       return PDC_SUCCESS;
>> +
>> +fail_dealloc:
>> +       dma_pool_free(pdcs->ring_pool, tx.vbase, tx.dmabase);
>> +done:
>> +       return err;
>> +}
>> +
>> +static void pdc_ring_free(struct pdc_state *pdcs)
>> +{
>> +       if (pdcs->tx_ring_alloc.vbase) {
>> +               dma_pool_free(pdcs->ring_pool, pdcs->tx_ring_alloc.vbase,
>> +                             pdcs->tx_ring_alloc.dmabase);
>> +               pdcs->tx_ring_alloc.vbase = NULL;
>> +       }
>> +
>> +       if (pdcs->rx_ring_alloc.vbase) {
>> +               dma_pool_free(pdcs->ring_pool, pdcs->rx_ring_alloc.vbase,
>> +                             pdcs->rx_ring_alloc.dmabase);
>> +               pdcs->rx_ring_alloc.vbase = NULL;
>> +       }
>> +}
>> +
>> +/**
>> + * pdc_send_data() - mailbox send_data function
>> + * @chan:      The mailbox channel on which the data is sent. The channel
>> + *              corresponds to a DMA ringset.
>> + * @data:      The mailbox message to be sent. The message must be a
>> + *              brcm_message structure.
>> + *
>> + * This function is registered as the send_data function for the mailbox
>> + * controller. From the destination scatterlist in the mailbox message, it
>> + * creates a sequence of receive descriptors in the rx ring. From the source
>> + * scatterlist, it creates a sequence of transmit descriptors in the tx ring.
>> + * After creating the descriptors, it writes the rx ptr and tx ptr registers to
>> + * initiate the DMA transfer.
>> + *
>> + * This function does the DMA map and unmap of the src and dst scatterlists in
>> + * the mailbox message.
>> + *
>> + * Return: 0 if successful
>> + *        -ENOTSUPP if the mailbox message is a type this driver does not
>> + *                     support
>> + *         < 0 if an error
>> + */
>> +static int pdc_send_data(struct mbox_chan *chan, void *data)
>> +{
>> +       struct pdc_state *pdcs = chan->con_priv;
>> +       struct device *dev = &pdcs->pdev->dev;
>> +       struct brcm_message *mssg = data;
>> +       int err = PDC_SUCCESS;
>> +       int src_nent;
>> +       int dst_nent;
>> +       int nent;
>> +
>> +       if (mssg->type != BRCM_MESSAGE_SPU)
>> +               return -ENOTSUPP;
>> +
>> +       src_nent = sg_nents(mssg->spu.src);
>> +       if (src_nent) {
>> +               nent = dma_map_sg(dev, mssg->spu.src, src_nent, DMA_TO_DEVICE);
>> +               if (nent == 0)
>> +                       return -EIO;
>> +       }
>> +
>> +       dst_nent = sg_nents(mssg->spu.dst);
>> +       if (dst_nent) {
>> +               nent = dma_map_sg(dev, mssg->spu.dst, dst_nent,
>> +                                 DMA_FROM_DEVICE);
>> +               if (nent == 0) {
>> +                       dma_unmap_sg(dev, mssg->spu.src, src_nent,
>> +                                    DMA_TO_DEVICE);
>> +                       return -EIO;
>> +               }
>> +       }
>> +
>> +       spin_lock(&pdcs->pdc_lock);
>> +
>> +       /* Create rx descriptors to SPU catch response */
>> +       err = pdc_rx_list_init(pdcs, mssg->spu.dst, mssg->ctx);
>> +       err |= pdc_rx_list_sg_add(pdcs, mssg->spu.dst);
>> +
>> +       /* Create tx descriptors to submit SPU request */
>> +       err |= pdc_tx_list_sg_add(pdcs, mssg->spu.src);
>> +       err |= pdc_tx_list_final(pdcs); /* initiate transfer */
>> +
>> +       spin_unlock(&pdcs->pdc_lock);
>> +
>> +       if (err)
>> +               dev_err(&pdcs->pdev->dev,
>> +                       "%s failed with error %d", __func__, err);
>> +
>> +       return err;
>> +}
>> +
>> +static int pdc_startup(struct mbox_chan *chan)
>> +{
>> +       return pdc_ring_init(chan->con_priv, PDC_RINGSET);
>> +}
>> +
>> +static void pdc_shutdown(struct mbox_chan *chan)
>> +{
>> +       struct pdc_state *pdcs = chan->con_priv;
>> +
>> +       if (pdcs)
>> +               dev_dbg(&pdcs->pdev->dev,
>> +                       "Shutdown mailbox channel for PDC %u", pdcs->pdc_idx);
>> +
>> +       pdc_ring_free(pdcs);
>> +}
>> +
>> +/**
>> + * pdc_hw_init() - Use the given initialization parameters to initialize the
>> + * state for one of the PDCs.
>> + * @dev:   device structure for PDC
>> + * @pdcs:  state of the PDC
>> + * @parms: parameter values to set
>> + */
>> +static
>> +void pdc_hw_init(struct device *dev, struct pdc_state *pdcs,
>> +                struct hw_init_parms *parms)
>>
> You could do without the dev and parms.

[WRR] Yes, I can get dev from pdcs. but parms is a stack variable in
the calling function and has to be passed, as I see it.

>
>> +{
>> +       struct dma64 *dma_reg;
>> +
>> +       int ringset = PDC_RINGSET;
>> +
>> +       dev_dbg(dev, "PDC %u initial values:", pdcs->pdc_idx);
>> +       dev_dbg(dev, "state structure:                   %p",
>> +               pdcs);
>> +       dev_dbg(dev, " - base physical addr of hw regs   %#llx",
>> +               parms->hw_pbase);
>> +       dev_dbg(dev, " - base virtual addr of hw regs    %p",
>> +               parms->hw_vbase);
>> +
>> +       /* initialize data structures */
>> +       pdcs->regs = (struct pdc_regs *)parms->hw_vbase;
>> +       pdcs->txregs_64 = (struct dma64_regs *)
>> +           (void *)(((u8 *)parms->hw_vbase) +
>> +                    PDC_TXREGS_OFFSET + (sizeof(struct dma64) * ringset));
>> +       pdcs->rxregs_64 = (struct dma64_regs *)
>> +           (void *)(((u8 *)parms->hw_vbase) +
>> +                    PDC_RXREGS_OFFSET + (sizeof(struct dma64) * ringset));
>> +
>> +       pdcs->ntxd = parms->ring_entries;
>> +       pdcs->nrxd = parms->ring_entries;
>> +       pdcs->ntxpost = parms->ring_entries - 1;
>> +       pdcs->nrxpost = parms->ring_entries - 1;
>> +       pdcs->regs->intmask = 0;
>> +
>> +       dma_reg = &pdcs->regs->dmaregs[ringset];
>> +       iowrite32(0, (void *)&dma_reg->dmaxmt.ptr);
>> +       iowrite32(0, (void *)&dma_reg->dmarcv.ptr);
>> +
>> +       iowrite32(PDC_TX_CTL, (void *)&dma_reg->dmaxmt.control);
>> +
>> +       iowrite32(PDC_RX_CTL + (pdcs->rx_status_len << 1),
>> +                 (void *)&dma_reg->dmarcv.control);
>> +
>> +       if (pdcs->pdc_resp_hdr_len == PDC_SPU2_RESP_HDR_LEN)
>> +               iowrite32(PDC_CKSUM_CTRL,
>> +                         pdcs->pdc_reg_vbase + PDC_CKSUM_CTRL_OFFSET);
>> +}
>> +
>> +/**
>> + * pdc_rx_buf_pool_create() - Pool of receive buffers used to catch the metadata
>> + * header returned with each response message.
>> + * @dev:  device structure
>> + * @pdcs: PDC state structure
>> + *
>> + * The metadata is not returned to the mailbox client. So the PDC driver
>> + * manages these buffers.
>> + *
>> + * Return: PDC_SUCCESS
>> + *         -ENOMEM if pool creation fails
>> + */
>> +static int pdc_rx_buf_pool_create(struct device *dev, struct pdc_state *pdcs)
>> +{
>> +       pdcs->pdc_resp_hdr_len = pdcs->rx_status_len;
>> +       if (pdcs->use_bcm_hdr)
>> +               pdcs->pdc_resp_hdr_len += BCM_HDR_LEN;
>> +
>> +       pdcs->rx_buf_pool = dma_pool_create("pdc rx bufs", dev,
>> +                                           pdcs->pdc_resp_hdr_len,
>> +                                           RX_BUF_ALIGN, 0);
>> +       if (!pdcs->rx_buf_pool)
>> +               return -ENOMEM;
>> +
>> +       return PDC_SUCCESS;
>> +}
>> +
>> +/**
>> + * pdc_interrupts_init() - Initialize the interrupt configuration for a PDC and
>> + * specify a threaded IRQ handler for deferred handling of interrupts outside of
>> + * interrupt context.
>> + * @pdev:   platform device
>> + * @pdcs:   PDC state
>> + *
>> + * Set the interrupt mask for transmit and receive done.
>> + * Set the lazy interrupt frame count to generate an interrupt for just one pkt.
>> + *
>> + * Return:  PDC_SUCCESS
>> + *          <0 if threaded irq request fails
>> + */
>> +static int pdc_interrupts_init(struct platform_device *pdev,
>> +                              struct pdc_state *pdcs)
>> +{
>> +       struct device *dev = &pdev->dev;
>> +       struct device_node *dn = pdev->dev.of_node;
>> +       int err;
>> +
>> +       pdcs->intstatus = 0;
>> +
>> +       /* interrupt configuration */
>> +       iowrite32(PDC_INTMASK, pdcs->pdc_reg_vbase + PDC_INTMASK_OFFSET);
>> +       iowrite32(PDC_LAZY_INT, pdcs->pdc_reg_vbase + PDC_RCVLAZY0_OFFSET);
>> +
>> +       /* read irq from device tree */
>> +       pdcs->pdc_irq = irq_of_parse_and_map(dn, 0);
>> +       dev_dbg(dev, "pdc device %s irq %u for pdcs %p",
>> +               dev_name(dev), pdcs->pdc_irq, pdcs);
>> +       err = devm_request_threaded_irq(dev, pdcs->pdc_irq,
>> +                                       pdc_irq_handler,
>> +                                       pdc_irq_thread, 0, dev_name(dev), pdcs);
>> +       if (err) {
>> +               dev_err(dev, "threaded tx IRQ %u request failed with err %d\n",
>> +                       pdcs->pdc_irq, err);
>> +               return err;
>> +       }
>> +       return PDC_SUCCESS;
>> +}
>> +
>> +static const struct mbox_chan_ops pdc_mbox_chan_ops = {
>> +       .send_data = pdc_send_data,
>> +       .startup = pdc_startup,
>> +       .shutdown = pdc_shutdown
>> +};
>> +
>> +/**
>> + * pdc_mb_init() - Initialize the mailbox controller.
>> + * @pdcs:  PDC state
>> + *
>> + * Each PDC is a mailbox controller. Each ringset is a mailbox channel. Kernel
>> + * driver only uses one ringset and thus one mb channel. PDC uses the transmit
>> + * complete interrupt to determine when a mailbox message has successfully been
>> + * transmitted.
>> + *
>> + * Return: 0 on success
>> + *         < 0 if there is an allocation or registration failure
>> + */
>> +static int pdc_mb_init(struct pdc_state *pdcs)
>> +{
>> +       struct device *dev = &pdcs->pdev->dev;
>> +       struct mbox_controller *mbc;
>> +       int chan_index;
>> +       int err;
>> +
>> +       mbc = &pdcs->mbc;
>> +       mbc->dev = dev;
>> +       mbc->ops = &pdc_mbox_chan_ops;
>> +       mbc->num_chans = 1;
>> +       mbc->chans = devm_kcalloc(dev, mbc->num_chans, sizeof(*mbc->chans),
>> +                                 GFP_KERNEL);
>> +       if (!mbc->chans)
>> +               return -ENOMEM;
>> +
>> +       mbc->txdone_irq = true;
>> +       mbc->txdone_poll = false;
>> +       for (chan_index = 0; chan_index < mbc->num_chans; chan_index++)
>> +               mbc->chans[chan_index].con_priv = pdcs;
>> +
>> +       /* Register mailbox controller */
>> +       err = mbox_controller_register(mbc);
>> +       if (err) {
>> +               dev_crit(dev,
>> +                        "Failed to register PDC mailbox controller. Error %d.",
>> +                        err);
>> +               return err;
>> +       }
>> +       return 0;
>> +}
>> +
>> +/**
>> + * pdc_dt_read() - Read application-specific data from device tree.
>> + * @pdev:  Platform device
>> + * @pdcs:  PDC state
>> + *
>> + * Reads the number of bytes of receive status that precede each received frame.
>> + * Reads whether transmit and received frames should be preceded by an 8-byte
>> + * BCM header.
>> + *
>> + * Return: 0 if successful
>> + *         -ENODEV if device not available
>> + */
>> +static int pdc_dt_read(struct platform_device *pdev, struct pdc_state *pdcs)
>> +{
>> +       struct device *dev = &pdev->dev;
>> +       struct device_node *dn = pdev->dev.of_node;
>> +       int err;
>> +
>> +       err = of_property_read_u32(dn, "brcm,rx-status-len",
>> +                                  &pdcs->rx_status_len);
>> +       if (err < 0)
>> +               dev_err(dev,
>> +                       "%s failed to get DMA receive status length from device tree",
>> +                       __func__);
>> +
>> +       pdcs->use_bcm_hdr = of_property_read_bool(dn, "brcm,use-bcm-hdr");
>> +
>> +       return 0;
>> +}
>> +
>> +/**
>> + * pdc_probe() - Probe function for PDC driver.
>> + * @pdev:   PDC platform device
>> + *
>> + * Reserve and map register regions defined in device tree.
>> + * Allocate and initialize tx and rx DMA rings.
>> + * Initialize a mailbox controller for each PDC.
>> + *
>> + * Return: 0 if successful
>> + *         < 0 if an error
>> + */
>> +static int pdc_probe(struct platform_device *pdev)
>> +{
>> +       int err = 0;
>> +       struct device *dev = &pdev->dev;
>> +       struct resource *pdc_regs;
>> +       struct pdc_state *pdcs;
>> +       struct hw_init_parms hw_parms;  /* params for initializing spu-dma */
>> +
>> +       /* PDC state for one SPU */
>> +       pdcs = devm_kzalloc(dev, sizeof(*pdcs), GFP_KERNEL);
>> +       if (!pdcs) {
>> +               err = -ENOMEM;
>> +               goto cleanup;
>> +       }
>> +
>> +       spin_lock_init(&pdcs->pdc_lock);
>> +       pdcs->pdev = pdev;
>> +       platform_set_drvdata(pdev, pdcs);
>> +       pdcs->pdc_idx = pdcg.num_spu;
>> +       pdcg.num_spu++;
>> +
>> +       err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32));
>> +       if (err) {
>> +               dev_warn(dev, "PDC device cannot perform DMA. Error %d.", err);
>> +               goto cleanup;
>> +       }
>> +
>> +       /* Create DMA pool for tx ring */
>> +       pdcs->ring_pool = dma_pool_create("pdc rings", dev, PDC_RING_SIZE,
>> +                                         RING_ALIGN, 0);
>> +       if (!pdcs->ring_pool) {
>> +               err = -ENOMEM;
>> +               goto cleanup;
>> +       }
>> +
>> +       err = pdc_dt_read(pdev, pdcs);
>> +       if (err)
>> +               goto cleanup_ring_pool;
>> +
>> +       pdc_regs = platform_get_resource(pdev, IORESOURCE_MEM, 0);
>> +       if (!pdc_regs) {
>> +               err = -ENODEV;
>> +               goto cleanup_ring_pool;
>> +       }
>> +       dev_dbg(dev, "PDC register region res.start = %#llx, res.end = %#llx",
>> +               pdc_regs->start, pdc_regs->end);
>> +
>> +       pdcs->pdc_reg_vbase = devm_ioremap_resource(&pdev->dev, pdc_regs);
>> +       if (IS_ERR(pdcs->pdc_reg_vbase)) {
>> +               err = PTR_ERR(pdcs->pdc_reg_vbase);
>> +               dev_err(&pdev->dev, "Failed to map registers: %d\n", err);
>> +               goto cleanup_ring_pool;
>> +       }
>> +
>> +       /* create rx buffer pool after dt read to know how big buffers are */
>> +       err = pdc_rx_buf_pool_create(dev, pdcs);
>          'dev' argument could be omitted.

[WRR] Yes, I can do that.

>
>> +       if (err)
>> +               goto cleanup_ring_pool;
>> +
>> +       hw_parms.hw_pbase = (dma_addr_t)(pdc_regs->start);
>> +       hw_parms.hw_vbase = pdcs->pdc_reg_vbase;
>> +
>> +       hw_parms.ring_entries = PDC_RING_ENTRIES;
>> +
>> +       pdc_hw_init(dev, pdcs, &hw_parms);
>      'dev' and 'hw_parms' argument could be omitted.

[WRR] Removed dev.

>
>> +
>> +       err = pdc_interrupts_init(pdev, pdcs);
>      'pdev' argument could be omitted.

[WRR] OK.

>
>
> thanks.

Re: [PATCH 2/3] mailbox: Add Broadcom PDC mailbox driver

[Jassi Brar]

On Tue, Jun 28, 2016 at 7:06 PM, Rob Rice <rob.rice@broadcom.com> wrote:

>>>  drivers/mailbox/Kconfig                 |    9 +
>>>  drivers/mailbox/Makefile                |    2 +
>>>  drivers/mailbox/mailbox-pdc/Makefile    |    8 +
>>>  drivers/mailbox/mailbox-pdc/pdc.c       | 1181 +++++++++++++++++++++++++++++++
>>>  drivers/mailbox/mailbox-pdc/pdc.h       |  303 ++++++++
>>>  drivers/mailbox/mailbox-pdc/pdc_debug.c |  101 +++
>>>  drivers/mailbox/mailbox-pdc/pdc_debug.h |   24 +
>>>
>> We already have mailbox api testing rig - mailbox-test, still if you
>> really need debugfs exposure please merge it in pdc.c especially when
>> pdc_debug.c is only 80lines and it always compiled and currently has
>> to export functions globally. Maybe fold everything in a single like
>> other platforms?
>
> [WRR] I'm using debugfs to display stats, not so much for testing the
> module. So I don't think mailbox-test is a substitute. But if you
> prefer, I'll move the stats code into pdc.c, along with the code from
> pdc.h and move pdc.c from the mailbox-pdc directory up to
> drivers/mailbox.
>
Yes, please merge them.

>>> +/**
>>> + * pdc_hw_init() - Use the given initialization parameters to initialize the
>>> + * state for one of the PDCs.
>>> + * @dev:   device structure for PDC
>>> + * @pdcs:  state of the PDC
>>> + * @parms: parameter values to set
>>> + */
>>> +static
>>> +void pdc_hw_init(struct device *dev, struct pdc_state *pdcs,
>>> +                struct hw_init_parms *parms)
>>>
>> You could do without the dev and parms.
>
> [WRR] Yes, I can get dev from pdcs. but parms is a stack variable in
> the calling function and has to be passed, as I see it.
>
hw_init_parms is invented only to be passed to pdc_hw_init from probe.
.ring_entries is PDC_RING_ENTRIES
.hw_pbase is unused
.hw_vbase is pdcs->pdc_reg_vbase

thanks.

[PATCH 3/3] arm64: dts: Add Broadcom Northstar2 device tree entries for PDC driver.

[Rob Rice]

From: Rob Rice <rrice@broadcom.com>

Add Broadcom Northstar2 SoC device tree entries for PDC driver.

Signed-off-by: Rob Rice <rob.rice@broadcom.com>
Reviewed-by: Ray Jui <ray.jui@broadcom.com>
Reviewed-by: Scott Branden <scott.branden@broadcom.com>
---
 arch/arm64/boot/dts/broadcom/ns2.dtsi | 36 +++++++++++++++++++++++++++++++++++
 1 file changed, 36 insertions(+)

diff --git a/arch/arm64/boot/dts/broadcom/ns2.dtsi b/arch/arm64/boot/dts/broadcom/ns2.dtsi
index ec68ec1..ba12702 100644
--- a/arch/arm64/boot/dts/broadcom/ns2.dtsi
+++ b/arch/arm64/boot/dts/broadcom/ns2.dtsi
@@ -191,6 +191,42 @@
 
 		#include "ns2-clock.dtsi"
 
+		pdc0: iproc-pdc0@612c0000 {
+			compatible = "brcm,iproc-pdc-mbox";
+			reg = <0x612c0000 0x445>;  /* PDC FS0 regs */
+			interrupts = <GIC_SPI 187 IRQ_TYPE_LEVEL_HIGH>;
+			#mbox-cells = <1>;
+			brcm,rx-status-len = <32>;
+			brcm,use-bcm-hdr;
+		};
+
+		pdc1: iproc-pdc1@612e0000 {
+			compatible = "brcm,iproc-pdc-mbox";
+			reg = <0x612e0000 0x445>;  /* PDC FS1 regs */
+			interrupts = <GIC_SPI 189 IRQ_TYPE_LEVEL_HIGH>;
+			#mbox-cells = <1>;
+			brcm,rx-status-len = <32>;
+			brcm,use-bcm-hdr;
+		};
+
+		pdc2: iproc-pdc2@61300000 {
+			compatible = "brcm,iproc-pdc-mbox";
+			reg = <0x61300000 0x445>;  /* PDC FS2 regs */
+			interrupts = <GIC_SPI 191 IRQ_TYPE_LEVEL_HIGH>;
+			#mbox-cells = <1>;
+			brcm,rx-status-len = <32>;
+			brcm,use-bcm-hdr;
+		};
+
+		pdc3: iproc-pdc3@61320000 {
+			compatible = "brcm,iproc-pdc-mbox";
+			reg = <0x61320000 0x445>;  /* PDC FS3 regs */
+			interrupts = <GIC_SPI 193 IRQ_TYPE_LEVEL_HIGH>;
+			#mbox-cells = <1>;
+			brcm,rx-status-len = <32>;
+			brcm,use-bcm-hdr;
+		};
+
 		dma0: dma@61360000 {
 			compatible = "arm,pl330", "arm,primecell";
 			reg = <0x61360000 0x1000>;
-- 
2.1.0