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

Rob Rice rob.rice at broadcom.com
Tue Jun 14 07:02:44 PDT 2016


From: Rob Rice <rrice at 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 at broadcom.com>
Reviewed-by: Scott Branden <scott.branden at broadcom.com>
Reviewed-by: Ray Jui <ray.jui at 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 at 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




More information about the linux-arm-kernel mailing list