LKML Archive on lore.kernel.org
 help / color / Atom feed
From: mgross@linux.intel.com
To: markgross@kernel.org, mgross@linux.intel.com, arnd@arndb.de,
	bp@suse.de, damien.lemoal@wdc.com, dragan.cvetic@xilinx.com,
	gregkh@linuxfoundation.org, corbet@lwn.net,
	leonard.crestez@nxp.com, palmerdabbelt@google.com,
	paul.walmsley@sifive.com, peng.fan@nxp.com, robh+dt@kernel.org,
	shawnguo@kernel.org, jassisinghbrar@gmail.com
Cc: linux-kernel@vger.kernel.org,
	Daniele Alessandrelli <daniele.alessandrelli@intel.com>
Subject: [PATCH v2 05/34] keembay-ipc: Add Keem Bay IPC module
Date: Fri,  8 Jan 2021 13:25:31 -0800
Message-ID: <20210108212600.36850-6-mgross@linux.intel.com> (raw)
In-Reply-To: <20210108212600.36850-1-mgross@linux.intel.com>

From: Daniele Alessandrelli <daniele.alessandrelli@intel.com>

On the Intel Movidius SoC code named Keem Bay, communication between the
Application Processor(AP) and the VPU is enabled by the Keem Bay
Inter-Processor
Communication (IPC) mechanism.

Add the driver for using Keem Bay IPC from within the Linux Kernel.

The IPC uses the following terminology:

- Node:    A processing entity that can use the IPC to communicate
           (currently, we just have two nodes, the AP and the VPU).

- Link:    Two nodes that can communicate over IPC form an IPC link
           (currently, we just have one link, the one formed by the AP
           and the VPU).

- Channel: An IPC link can provide multiple IPC channels. IPC channels
           allow communication multiplexing, i.e., the same IPC link can
           be used by different applications for different
           communications. Each channel is identified by a channel ID,
           which must be unique within a single IPC link. Channels are
           divided in two categories, High-Speed (HS) channels and
           General-Purpose (GP) channels. HS channels have higher
           priority over GP channels.

The Keem Bay IPC mechanism is built on top of the VPU IPC mailbox, which
allows the AP and the VPU to exchange 32-bit messages. Specifically, the
IPC uses shared memory (shared between the AP and the VPU) to allocate
IPC packets and then exchanges them using the VPU IPC mailbox (the
32-bit physical address of the packet is passed as a message to the VPU
IPC mailbox).

IPC packets have a fixed structure containing the (VPU) physical address
of the payload (which must be located in shared memory too) as well as
other information (payload size, IPC channel ID, etc.).

Each IPC node (i.e., both the AP and the VPU) has its own reserved
memory region (in shared memory) from which it instantiates its own pool
of IPC packets.  When instantiated, IPC packets are marked as free. When
the node needs to send an IPC message, it gets the first free packet it
finds (from its own pool), marks it as allocated (used), and transfer
its physical address to the destination node using the VPU IPC mailbox.
The destination node uses the received physical address to access the
IPC packet, process the packet, and, once done with it, marks it as free
(so that the sender can reuse it).

Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Palmer Dabbelt <palmerdabbelt@google.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Damien Le Moal <damien.lemoal@wdc.com>
Cc: Peng Fan <peng.fan@nxp.com>
Cc: Shawn Guo <shawnguo@kernel.org>
Cc: Leonard Crestez <leonard.crestez@nxp.com>
Reviewed-by: Mark Gross <mgross@linux.intel.com>
Signed-off-by: Daniele Alessandrelli <daniele.alessandrelli@intel.com>
---
 MAINTAINERS                           |    8 +
 drivers/soc/Kconfig                   |    1 +
 drivers/soc/Makefile                  |    1 +
 drivers/soc/intel/Kconfig             |   18 +
 drivers/soc/intel/Makefile            |    4 +
 drivers/soc/intel/keembay-ipc.c       | 1364 +++++++++++++++++++++++++
 include/linux/soc/intel/keembay-ipc.h |   30 +
 7 files changed, 1426 insertions(+)
 create mode 100644 drivers/soc/intel/Kconfig
 create mode 100644 drivers/soc/intel/Makefile
 create mode 100644 drivers/soc/intel/keembay-ipc.c
 create mode 100644 include/linux/soc/intel/keembay-ipc.h

diff --git a/MAINTAINERS b/MAINTAINERS
index 496baaf0c754..422047edbf0c 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -9062,6 +9062,14 @@ F:	drivers/crypto/keembay/keembay-ocs-aes-core.c
 F:	drivers/crypto/keembay/ocs-aes.c
 F:	drivers/crypto/keembay/ocs-aes.h
 
+INTEL KEEM BAY IPC DRIVER
+M:	Daniele Alessandrelli <daniele.alessandrelli@intel.com>
+M:	Mark Gross <mgross@linux.intel.com>
+S:	Supported
+F:	Documentation/devicetree/bindings/soc/intel/intel,keembay-ipc.yaml
+F:	drivers/soc/intel/keembay-ipc.c
+F:	include/linux/soc/intel/keembay-ipc.h
+
 INTEL MANAGEMENT ENGINE (mei)
 M:	Tomas Winkler <tomas.winkler@intel.com>
 L:	linux-kernel@vger.kernel.org
diff --git a/drivers/soc/Kconfig b/drivers/soc/Kconfig
index d097d070f579..b9d69a1eedc7 100644
--- a/drivers/soc/Kconfig
+++ b/drivers/soc/Kconfig
@@ -8,6 +8,7 @@ source "drivers/soc/atmel/Kconfig"
 source "drivers/soc/bcm/Kconfig"
 source "drivers/soc/fsl/Kconfig"
 source "drivers/soc/imx/Kconfig"
+source "drivers/soc/intel/Kconfig"
 source "drivers/soc/ixp4xx/Kconfig"
 source "drivers/soc/litex/Kconfig"
 source "drivers/soc/mediatek/Kconfig"
diff --git a/drivers/soc/Makefile b/drivers/soc/Makefile
index 699b758d28e4..1a6c00d2e32e 100644
--- a/drivers/soc/Makefile
+++ b/drivers/soc/Makefile
@@ -12,6 +12,7 @@ obj-$(CONFIG_MACH_DOVE)		+= dove/
 obj-y				+= fsl/
 obj-$(CONFIG_ARCH_GEMINI)	+= gemini/
 obj-y				+= imx/
+obj-y				+= intel/
 obj-$(CONFIG_ARCH_IXP4XX)	+= ixp4xx/
 obj-$(CONFIG_SOC_XWAY)		+= lantiq/
 obj-$(CONFIG_LITEX_SOC_CONTROLLER) += litex/
diff --git a/drivers/soc/intel/Kconfig b/drivers/soc/intel/Kconfig
new file mode 100644
index 000000000000..a575e31e47b4
--- /dev/null
+++ b/drivers/soc/intel/Kconfig
@@ -0,0 +1,18 @@
+# SPDX-License-Identifier: GPL-2.0-only
+#
+# Keem Bay SoC drivers
+#
+
+menu "Intel SoC drivers"
+
+config KEEMBAY_IPC
+	tristate "Support for Intel Keem Bay IPC"
+	depends on INTEL_VPU_IPC_MBOX
+	depends on ARCH_KEEMBAY || (ARM64 && COMPILE_TEST)
+	help
+	  Keem Bay IPC enables communication between the Keem Bay CPU
+	  Sub-System (CSS) and the Keem Bay Media Sub-System (MSS).
+
+	  Select this if you are compiling the Kernel for an Intel SoC that
+	  includes the Intel Vision Processing Unit (VPU) such as Keem Bay.
+endmenu
diff --git a/drivers/soc/intel/Makefile b/drivers/soc/intel/Makefile
new file mode 100644
index 000000000000..ecf0246e7822
--- /dev/null
+++ b/drivers/soc/intel/Makefile
@@ -0,0 +1,4 @@
+#
+# Makefile for Keem Bay IPC Linux driver
+#
+obj-$(CONFIG_KEEMBAY_IPC) += keembay-ipc.o
diff --git a/drivers/soc/intel/keembay-ipc.c b/drivers/soc/intel/keembay-ipc.c
new file mode 100644
index 000000000000..70a8d8b1f9a8
--- /dev/null
+++ b/drivers/soc/intel/keembay-ipc.c
@@ -0,0 +1,1364 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Keem Bay IPC Driver.
+ *
+ * Copyright (C) 2018-2020 Intel Corporation
+ *
+ * This driver implements the VPU Inter-Processor Communication (IPC) mechanism
+ * which enables communication between the Application Processor (AP), running
+ * Linux, and the VPU, running a proprietary VPU firmware.
+ *
+ * The IPC uses the following terminology:
+ *
+ * - Node:    A processing entity that can use the IPC to communicate
+ *	      (currently, we just have two nodes, the AP and the VPU).
+ *
+ * - Link:    Two nodes that can communicate over IPC form an IPC link
+ *	      (currently, we just have one link, the one formed by the AP and
+ *	      the VPU).
+ *
+ * - Channel: An IPC link can provide multiple IPC channels. IPC channels allow
+ *            communication multiplexing, i.e., the same IPC link can be used
+ *            by different applications for different communications. Each
+ *            channel is identified by a channel ID, which must be unique
+ *            within a single IPC link. Channels are divided in two categories,
+ *            High-Speed (HS) channels and General-Purpose (GP) channels.
+ *            HS channels have higher priority over GP channels.
+ *
+ * The VPU IPC mechanism is built on top of the VPU IPC mailbox, which allows
+ * the AP and the VPU to exchange 32-bit messages. Specifically, the VPU IPC
+ * mechanism uses shared memory (shared between the AP and the VPU) to allocate
+ * IPC packets and then exchanges them using the VPU IPC mailbox (the 32-bit
+ * physical address of the packet is passed as a message to the VPU IPC
+ * mailbox).
+ *
+ * IPC packets have a fixed structure containing the (VPU) physical address of
+ * the payload (which must be located in shared memory too) as well as other
+ * information (payload size, IPC channel ID, etc.).
+ *
+ * Each IPC node (i.e., both the AP and the VPU) has its own reserved memory
+ * region (in shared memory) from which it instantiates its own pool of IPC
+ * packets.  When instantiated, IPC packets are marked as free. When the node
+ * needs to send an IPC message, it gets the first free packet it finds (from
+ * its own pool), marks it as allocated (used), and transfer its physical
+ * address to the destination node using the VPU IPC mailbox. The destination
+ * node uses the received physical address to access the IPC packet, process
+ * the packet, and, once done with it, marks it as free (so that the sender can
+ * reuse it).
+ *
+ * Note: Keem Bay IPC is not based on RPMsg, since VPU HW/FW does not support
+ * Virtio and Virtqueues.
+ */
+
+#include <linux/circ_buf.h>
+#include <linux/completion.h>
+#include <linux/dma-mapping.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/jiffies.h>
+#include <linux/kernel.h>
+#include <linux/kthread.h>
+#include <linux/mailbox_client.h>
+#include <linux/mailbox_controller.h> /* Needed for MBOX_TX_QUEUE_LEN. */
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/of_address.h>
+#include <linux/of_device.h>
+#include <linux/of_irq.h>
+#include <linux/of_reserved_mem.h>
+#include <linux/platform_device.h>
+#include <linux/slab.h>
+#include <linux/wait.h>
+
+#include <linux/soc/intel/keembay-ipc.h>
+
+#define DRV_NAME		"keembay-ipc"
+
+/* The alignment to be used for IPC Packets and IPC Data. */
+#define KMB_IPC_ALIGNMENT	64
+
+/* Maximum number of channels per link. */
+#define KMB_IPC_MAX_CHANNELS	1024
+
+/* The number of high-speed channels per link. */
+#define KMB_IPC_NUM_HS_CHANNELS	10
+
+/*
+ * This is used as index for retrieving reserved memory from the device
+ * tree.
+ */
+#define RSVD_MEM_IDX_CPU_PKTS	0
+#define RSVD_MEM_IDX_VPU_PKTS	1
+
+/* The possible states of an IPC packet. */
+enum {
+	/*
+	 * KMB_IPC_PKT_FREE must be set to 0 to ensure that packets can be
+	 * initialized with memset(&buf, 0, sizeof(buf)).
+	 */
+	KMB_IPC_PKT_FREE = 0,
+	KMB_IPC_PKT_ALLOCATED,
+};
+
+/**
+ * struct kmb_ipc_pkt - The IPC packet structure.
+ * @data_addr:	The address where the IPC payload is located; NOTE: this is a
+ *		VPU address (not a CPU one).
+ * @data_size:	The size of the payload.
+ * @channel:	The channel used.
+ * @src_node:	The Node ID of the sender.
+ * @dst_node:	The Node ID of the intended receiver.
+ * @status:	Either free or allocated.
+ */
+struct kmb_ipc_pkt {
+	u32	data_addr;
+	u32	data_size;
+	u16	channel;
+	u8	src_node;
+	u8	dst_node;
+	u8	status;
+} __packed __aligned(KMB_IPC_ALIGNMENT);
+
+/**
+ * struct ipc_pkt_mem - IPC Packet Memory Region.
+ * @dev:	Child device managing the memory region.
+ * @vaddr:	The virtual address of the memory region.
+ * @dma_handle:	The VPU address of the memory region.
+ * @size:	The size of the memory region.
+ */
+struct ipc_pkt_mem {
+	struct device	*dev;
+	void		*vaddr;
+	dma_addr_t	dma_handle;
+	size_t		size;
+};
+
+/**
+ * struct ipc_pkt_pool - IPC packet pool.
+ * @packets:	Pointer to the array of packets.
+ * @buf_cnt:	Pool size (i.e., number of packets).
+ * @idx:	Current index.
+ */
+struct ipc_pkt_pool {
+	struct kmb_ipc_pkt	*packets;
+	size_t			buf_cnt;
+	size_t			idx;
+};
+
+/**
+ * struct ipc_chan - IPC channel.
+ * @rx_data_list:	The list of incoming messages.
+ * @rx_lock:		The lock for modifying the rx_data_list.
+ * @rx_wait_queue:	The wait queue for RX Data (recv() waits on it).
+ * @closing:		Closing flag, set when the channel is being closed.
+ */
+struct ipc_chan {
+	struct list_head	rx_data_list;
+	spinlock_t		rx_lock; /* Protects 'rx_data_list'. */
+	wait_queue_head_t	rx_wait_queue;
+	bool			closing;
+};
+
+/**
+ * struct tx_data - Element of a TX queue.
+ * @list:	The list head used to concatenate TX data elements.
+ * @vpu_addr:	The VPU address of the data to be transferred.
+ * @size:	The size of the data to be transferred.
+ * @chan_id:	The channel to be used for the transfer.
+ * @dst_node:	The destination node.
+ * @retv:	The result of the transfer.
+ * @tx_done:	The completion struct used by the sender to wait for the
+ *		transfer to complete.
+ * @entry:	The IPC packet VPU address to be sent to the VPU (this field is
+ *		set by tx_data_send() and used by ipc_mbox_tx_done() to get a
+ *		reference to the associated tx_data struct).
+ */
+struct tx_data {
+	struct	list_head list;
+	u32	vpu_addr;
+	u32	size;
+	u16	chan_id;
+	u8	dst_node;
+	int	retv;
+	struct	completion tx_done;
+	u32	entry;
+};
+
+/**
+ * struct tx_queue - The TX queue structure.
+ * @tx_data_list: The list of pending TX data on this queue.
+ * @lock:	  The lock protecting the TX data list.
+ */
+struct tx_queue {
+	struct list_head	tx_data_list;
+	spinlock_t		lock;	/* Protects tx_data_list. */
+};
+
+/**
+ * struct ipc_link - An IPC link.
+ * @mbox_cl:	   The mailbox client associated with this link.
+ * @mbox_chan:	   The mailbox channel associated with this link.
+ * @mbox_tx_queue: The completion used to avoid overflowing the mailbox
+ *		   framework queue (MBOX_TX_QUEUE_LEN), which will result in
+ *		   IPC packets being dropped.
+ * @channels:	   The channels associated with this link (the pointers to the
+ *		   channels are RCU-protected).
+ * @chan_lock:	   The lock for modifying the channels array.
+ * @srcu_sp:	   The Sleepable RCU structs associated with the link's
+ *		   channels.
+ * @tx_queues:	   The TX queues for this link (1 queue for each high-speed
+ *		   channels + 1 shared among all the general-purpose channels).
+ * @tx_qidx:	   The index of the next tx_queue to be check for outgoing data.
+ * @tx_queued:	   The TX completion used to signal when new TX data is pending.
+ * @tx_thread:	   The TX thread.
+ * @tx_stopping:   Flag signaling that the IPC Link is being closed.
+ */
+struct ipc_link {
+	struct mbox_client	mbox_cl;
+	struct mbox_chan	*mbox_chan;
+	struct completion       mbox_tx_queue;
+	struct ipc_chan __rcu	*channels[KMB_IPC_MAX_CHANNELS];
+	spinlock_t		chan_lock; /* Protects 'channels'. */
+	struct srcu_struct	srcu_sp[KMB_IPC_MAX_CHANNELS];
+	struct tx_queue		tx_queues[KMB_IPC_NUM_HS_CHANNELS + 1];
+	int			tx_qidx;
+	struct completion	tx_queued;
+	struct task_struct	*tx_thread;
+	bool			tx_stopping;
+};
+
+/**
+ * struct keembay_ipc_dev - IPC private data.
+ *
+ * @plat_dev: Platform device for this driver.
+ * @cpu_ipc_mem:	    Local IPC Packet memory region.
+ * @vpu_ipc_mem:	    Remove IPC Packet memory region.
+ * @ipc_pkt_pool:	    The pool of IPC packets.
+ * @vpu_link:		    The CPU-VPU link.
+ */
+struct keembay_ipc_dev {
+	struct platform_device	*plat_dev;
+	struct ipc_pkt_mem	cpu_ipc_mem;
+	struct ipc_pkt_mem	vpu_ipc_mem;
+	struct ipc_pkt_pool	ipc_pkt_pool;
+	struct ipc_link		vpu_link;
+};
+
+/**
+ * struct rx_data - RX Data Descriptor.
+ * @list:		List head for creating a list of rx_data elements.
+ * @data_vpu_addr:	The VPU address of the received data.
+ * @data_size:		The size of the received data.
+ *
+ * Instances of this struct are meant to be inserted in the RX Data queue
+ * (list) associated with each channel.
+ */
+struct rx_data {
+	struct list_head	list;
+	u32			data_vpu_addr;
+	u32			data_size;
+};
+
+/* Forward declaration of TX thread function. */
+static int tx_thread_fn(void *ptr);
+
+/* Forward declaration of mailbox client callbacks. */
+static void ipc_mbox_rx_callback(struct mbox_client *cl, void *msg);
+static void ipc_mbox_tx_done(struct mbox_client *cl, void *mssg, int r);
+
+/*
+ * Functions related to reserved-memory sub-devices.
+ */
+
+/*
+ * init_ipc_rsvd_mem() - Init the specified IPC reserved memory.
+ * @dev:	The IPC device for which the memory will be initialized.
+ * @mem:	Where to stored information about the initialized memory.
+ * @mem_name:	The name of this IPC memory.
+ * @mem_idx:	The index of the memory to initialize.
+ *
+ * Create a child device for 'dev' and use it to initialize the reserved
+ * memory region specified in the device tree at index 'mem_idx'.
+ * Also allocate DMA memory from the initialized memory region.
+ *
+ * Return:	0 on success, negative error code otherwise.
+ */
+static int init_ipc_rsvd_mem(struct device *dev, struct ipc_pkt_mem *mem,
+			     const char *mem_name, unsigned int mem_idx)
+{
+	struct device *mem_dev;
+	struct device_node *np;
+	struct resource mem_res;
+	dma_addr_t dma_handle;
+	size_t mem_size;
+	void *vaddr;
+	int rc;
+
+	/* Create a child device (of dev) to own the reserved memory. */
+	mem_dev = devm_kzalloc(dev, sizeof(struct device), GFP_KERNEL);
+	if (!mem_dev)
+		return -ENOMEM;
+
+	device_initialize(mem_dev);
+	dev_set_name(mem_dev, "%s:%s", dev_name(dev), mem_name);
+	mem_dev->parent = dev;
+	mem_dev->dma_mask = dev->dma_mask;
+	mem_dev->coherent_dma_mask = dev->coherent_dma_mask;
+	mem_dev->release = of_reserved_mem_device_release;
+
+	/* Set up DMA configuration using information from parent's DT node. */
+	rc = of_dma_configure(mem_dev, dev->of_node, true);
+	if (rc)
+		goto err_add;
+
+	rc = device_add(mem_dev);
+	if (rc)
+		goto err_add;
+
+	/* Initialized the device reserved memory region. */
+	rc = of_reserved_mem_device_init_by_idx(mem_dev, dev->of_node, mem_idx);
+	if (rc) {
+		dev_err(dev, "Couldn't get reserved memory with idx = %d, %d\n",
+			mem_idx, rc);
+		goto err_post_add;
+	}
+
+	/* Find out the size of the memory region. */
+	np = of_parse_phandle(dev->of_node, "memory-region", mem_idx);
+	if (!np) {
+		dev_err(dev, "Couldn't find memory-region %d\n", mem_idx);
+		rc = -EINVAL;
+		goto err_post_add;
+	}
+	rc = of_address_to_resource(np, 0, &mem_res);
+	if (rc) {
+		dev_err(dev, "Couldn't map address to resource %d\n", mem_idx);
+		goto err_post_add;
+	}
+	mem_size = resource_size(&mem_res);
+
+	/* Allocate memory from the reserved memory regions */
+	vaddr = dmam_alloc_coherent(mem_dev, mem_size, &dma_handle, GFP_KERNEL);
+	if (!vaddr) {
+		dev_err(mem_dev, "Failed to allocate from reserved memory.\n");
+		rc = -ENOMEM;
+		goto err_post_add;
+	}
+
+	mem->dev = mem_dev;
+	mem->vaddr = vaddr;
+	mem->dma_handle = dma_handle;
+	mem->size = mem_size;
+
+	return 0;
+
+err_post_add:
+	device_del(mem_dev);
+err_add:
+	put_device(mem_dev);
+	return rc;
+}
+
+/*
+ * IPC internal functions.
+ */
+
+/**
+ * channel_close() - Close a channel and return whether it was open or not.
+ * @link:	The link the channel belongs to.
+ * @chan_id:	The channel ID of the channel to close.
+ *
+ * Return:	0 if the channel was already closed, 1 otherwise.
+ */
+static int channel_close(struct ipc_link *link, u16 chan_id)
+{
+	struct ipc_chan *chan;
+	struct rx_data *pos, *nxt;
+
+	/* Get channel from channel array. */
+	spin_lock(&link->chan_lock);
+	chan = rcu_dereference_protected(link->channels[chan_id],
+					 lockdep_is_held(&link->chan_lock));
+
+	/* If channel is already NULL, we are done. */
+	if (!chan) {
+		spin_unlock(&link->chan_lock);
+		return 0;
+	}
+
+	/* Otherwise remove it from the 'channels' array. */
+	RCU_INIT_POINTER(link->channels[chan_id], NULL);
+	spin_unlock(&link->chan_lock);
+
+	/* Set closing flag and wake up user threads waiting on recv(). */
+	chan->closing = true;
+	wake_up_all(&chan->rx_wait_queue);
+
+	/* Wait for channel users to drop the reference to the old channel. */
+	synchronize_srcu(&link->srcu_sp[chan_id]);
+
+	/* Free channel memory (rx_data queue and channel struct). */
+	/*
+	 * No need to get chan->rx_lock as we know that nobody is using the
+	 * channel at this point.
+	 */
+	list_for_each_entry_safe(pos, nxt, &chan->rx_data_list, list) {
+		list_del(&pos->list);
+		kfree(pos);
+	}
+	kfree(chan);
+
+	return 1;
+}
+
+/**
+ * ipc_pkt_tx_alloc() - Allocate an IPC packet to be used for TX.
+ * @pool:  The IPC packet pool from which the IPC packet will be allocated.
+ *
+ * Return: The pointer to the allocated packet, or NULL if allocation fails.
+ */
+static struct kmb_ipc_pkt *ipc_pkt_tx_alloc(struct ipc_pkt_pool *pool)
+{
+	struct kmb_ipc_pkt *buf;
+	int i;
+
+	/*
+	 * Look for a free packet starting from the last index (pointing to the
+	 * next packet after the last allocated one) and potentially going
+	 * through all the packets in the pool.
+	 */
+	for (i = 0; i < pool->buf_cnt; i++) {
+		/*
+		 * Get reference to current packet and increment index (for
+		 * next iteration or function call).
+		 */
+		buf = &pool->packets[pool->idx++];
+		if (pool->idx == pool->buf_cnt)
+			pool->idx = 0;
+
+		/* Use current packet if free. */
+		if (buf->status == KMB_IPC_PKT_FREE) {
+			buf->status = KMB_IPC_PKT_ALLOCATED;
+			return buf;
+		}
+	}
+
+	/* We went through all the packets and found none free. */
+	return NULL;
+}
+
+/**
+ * init_ipc_pkt_pool() - Init the CPU IPC Packet Pool.
+ * @ipc_dev:	The IPC device the pool belongs to.
+ *
+ * Set up the IPC Packet Pool to be used for allocating TX packets.
+ *
+ * The pool uses the CPU IPC Packet memory previously allocated.
+ *
+ * Return:	0 on success, negative error code otherwise.
+ */
+static int init_ipc_pkt_pool(struct keembay_ipc_dev *ipc_dev)
+{
+	struct ipc_pkt_mem *mem = &ipc_dev->cpu_ipc_mem;
+
+	ipc_dev->ipc_pkt_pool.buf_cnt = mem->size / sizeof(struct kmb_ipc_pkt);
+
+	/* Fail if we end up having a pool of 0 packets. */
+	if (ipc_dev->ipc_pkt_pool.buf_cnt == 0)
+		return -ENOMEM;
+	/*
+	 * Set reserved memory to 0 to initialize the IPC Packet array
+	 * (ipc_pkt_pool.packets); this works because the value of
+	 * KMB_IPC_PKT_FREE is 0.
+	 */
+	memset(mem->vaddr, 0, mem->size);
+	ipc_dev->ipc_pkt_pool.packets = mem->vaddr;
+	ipc_dev->ipc_pkt_pool.idx = 0;
+
+	return 0;
+}
+
+/*
+ * ipc_link_init() - Initialize CPU/VPU IPC link.
+ * @ipc_dev: The IPC device the link belongs to.
+ *
+ * This function is expected to be called during probing.
+ *
+ * Return: 0 on success, negative error code otherwise.
+ */
+static int ipc_link_init(struct keembay_ipc_dev *ipc_dev)
+{
+	struct ipc_link *link = &ipc_dev->vpu_link;
+	struct tx_queue *queue;
+	int i;
+
+	/* Init mailbox client. */
+	link->mbox_cl.dev = &ipc_dev->plat_dev->dev;
+	link->mbox_cl.tx_block = false;
+	link->mbox_cl.tx_tout = 0;
+	link->mbox_cl.knows_txdone = false;
+	link->mbox_cl.rx_callback = ipc_mbox_rx_callback;
+	link->mbox_cl.tx_prepare = NULL;
+	link->mbox_cl.tx_done = ipc_mbox_tx_done;
+
+	/* Init completion keeping track of free slots in mbox tx queue. */
+	init_completion(&link->mbox_tx_queue);
+	for (i = 0; i < MBOX_TX_QUEUE_LEN; i++)
+		complete(&link->mbox_tx_queue);
+
+	/* Request mailbox channel */
+	link->mbox_chan = mbox_request_channel(&link->mbox_cl, 0);
+	if (IS_ERR(link->mbox_chan))
+		return PTR_ERR(link->mbox_chan);
+
+	spin_lock_init(&link->chan_lock);
+	for (i = 0; i < ARRAY_SIZE(link->srcu_sp); i++)
+		init_srcu_struct(&link->srcu_sp[i]);
+	memset(link->channels, 0, sizeof(link->channels));
+
+	/* Init TX queues. */
+	for (i = 0; i < ARRAY_SIZE(link->tx_queues); i++) {
+		queue = &link->tx_queues[i];
+		INIT_LIST_HEAD(&queue->tx_data_list);
+		spin_lock_init(&queue->lock);
+	}
+	link->tx_qidx = 0;
+	link->tx_stopping = false;
+	init_completion(&link->tx_queued);
+
+	/* Start TX thread. */
+	link->tx_thread = kthread_run(tx_thread_fn, ipc_dev,
+				      "kmb_ipc_tx_thread");
+	if (IS_ERR(link->tx_thread)) {
+		mbox_free_channel(link->mbox_chan);
+		return PTR_ERR(link->tx_thread);
+	}
+
+	return 0;
+}
+
+/**
+ * ipc_link_deinit() - De-initialize CPU/VPU IPC link.
+ * @ipc_dev:	The IPC device the link belongs to.
+ */
+static void ipc_link_deinit(struct keembay_ipc_dev *ipc_dev)
+{
+	struct ipc_link *link = &ipc_dev->vpu_link;
+	struct tx_queue *queue;
+	struct tx_data *pos, *nxt;
+	int i;
+
+	/* Close all channels. */
+	for (i = 0; i < ARRAY_SIZE(link->channels); i++)
+		channel_close(link, i);
+
+	/* Stop TX Thread. */
+	link->tx_stopping = true;
+	complete(&link->tx_queued);
+	kthread_stop(link->tx_thread);
+
+	/* Flush all TX queue. */
+	for (i = 0; i < ARRAY_SIZE(link->tx_queues); i++) {
+		queue = &link->tx_queues[i];
+		list_for_each_entry_safe(pos, nxt, &queue->tx_data_list, list) {
+			list_del(&pos->list);
+			pos->retv = -EPIPE;
+			complete(&pos->tx_done);
+		}
+	}
+
+	mbox_free_channel(link->mbox_chan);
+}
+
+/* Driver probing. */
+static int kmb_ipc_probe(struct platform_device *pdev)
+{
+	int rc;
+	struct keembay_ipc_dev *ipc_dev;
+	struct device *dev = &pdev->dev;
+
+	ipc_dev = devm_kzalloc(dev, sizeof(*ipc_dev), GFP_KERNEL);
+	if (!ipc_dev)
+		return -ENOMEM;
+
+	ipc_dev->plat_dev = pdev;
+
+	/* Init the memory used for CPU packets. */
+	rc = init_ipc_rsvd_mem(dev, &ipc_dev->cpu_ipc_mem,
+			       "ipc_cpu_rsvd_mem", RSVD_MEM_IDX_CPU_PKTS);
+	if (rc) {
+		dev_err(dev, "Failed to set up CPU reserved memory.\n");
+		return rc;
+	}
+
+	/* Init the memory used for VPU packets. */
+	rc = init_ipc_rsvd_mem(dev, &ipc_dev->vpu_ipc_mem,
+			       "ipc_vpu_rsvd_mem", RSVD_MEM_IDX_VPU_PKTS);
+	if (rc) {
+		dev_err(dev, "Failed to set up VPU reserved memory.\n");
+		device_unregister(ipc_dev->cpu_ipc_mem.dev);
+		return rc;
+	}
+
+	/* Init the pool of IPC packets to be used to TX. */
+	rc = init_ipc_pkt_pool(ipc_dev);
+	if (rc)
+		goto err_post_rsvd_mem;
+
+	/* Init the only link we have (CPU -> VPU). */
+	rc = ipc_link_init(ipc_dev);
+	if (rc)
+		goto err_post_rsvd_mem;
+
+	platform_set_drvdata(pdev, ipc_dev);
+
+	return 0;
+
+err_post_rsvd_mem:
+	device_unregister(ipc_dev->cpu_ipc_mem.dev);
+	device_unregister(ipc_dev->vpu_ipc_mem.dev);
+
+	return rc;
+}
+
+/* Driver removal. */
+static int kmb_ipc_remove(struct platform_device *pdev)
+{
+	struct keembay_ipc_dev *ipc_dev = platform_get_drvdata(pdev);
+
+	ipc_link_deinit(ipc_dev);
+
+	/*
+	 * No need to de-alloc IPC memory (cpu_ipc_mem and vpu_ipc_mem)
+	 * since it was allocated with dmam_alloc_coherent().
+	 */
+
+	device_unregister(ipc_dev->cpu_ipc_mem.dev);
+	device_unregister(ipc_dev->vpu_ipc_mem.dev);
+
+	return 0;
+}
+
+/*
+ * ipc_vpu_to_virt() - Convert a VPU address to a CPU virtual address.
+ *
+ * @ipc_mem:  The IPC memory region where the VPU address is expected to be
+ *	      mapped to.
+ * @vpu_addr: The VPU address to be converted to a virtual one.
+ *
+ * The VPU can map the DDR differently than the CPU. This functions converts
+ * VPU addresses to CPU virtual addresses.
+ *
+ * Return: The corresponding CPU virtual address, or NULL if the VPU address
+ *	   is not in the expected memory range.
+ */
+static void *ipc_vpu_to_virt(const struct ipc_pkt_mem *ipc_mem,
+			     u32 vpu_addr)
+{
+	if (unlikely(vpu_addr < ipc_mem->dma_handle ||
+		     vpu_addr >= (ipc_mem->dma_handle + ipc_mem->size)))
+		return NULL;
+
+	/* Cast to (u8 *) needed since void pointer arithmetic is undefined. */
+	return (u8 *)ipc_mem->vaddr + (vpu_addr - ipc_mem->dma_handle);
+}
+
+/*
+ * ipc_virt_to_vpu() - Convert an CPU virtual address to a VPU address.
+ * @ipc_mem:  [in]  The IPC memory region where the VPU address is expected to
+ *		    be mapped to.
+ * @vaddr:    [in]  The CPU virtual address to be converted to a VPU one.
+ * @vpu_addr: [out] Where to store the computed VPU address.
+ *
+ * The VPU can map the DDR differently than the CPU. This functions converts
+ * CPU virtual addresses to VPU virtual addresses.
+ *
+ * Return: 0 on success, negative error code otherwise.
+ */
+static int ipc_virt_to_vpu(struct ipc_pkt_mem *ipc_mem, void *vaddr,
+			   u32 *vpu_addr)
+{
+	if (unlikely((ipc_mem->dma_handle + ipc_mem->size) > 0xFFFFFFFF))
+		return -EINVAL;
+
+	/* Cast to (u8 *) needed since void pointer arithmetic is undefined. */
+	if (unlikely(vaddr < ipc_mem->vaddr ||
+		     (u8 *)vaddr >= ((u8 *)ipc_mem->vaddr + ipc_mem->size)))
+		return -EINVAL;
+
+	*vpu_addr = ipc_mem->dma_handle + (vaddr - ipc_mem->vaddr);
+
+	return 0;
+}
+
+/**
+ * ipc_mbox_rx_callback() - Process a FIFO entry coming from IPC mailbox.
+ * @cl:		The mailbox client.
+ * @msg:	The FIFO entry to process.
+ *
+ * This function performs the following tasks:
+ * - Check the source node id.
+ * - Process the IPC packet (locate it, validate it, read data info, release
+ *   packet).
+ * - Add an RX Data descriptor (data ptr and data size) to the channel RX queue.
+ */
+static void ipc_mbox_rx_callback(struct mbox_client *cl, void *msg)
+{
+	struct device *dev = cl->dev;
+	struct ipc_link *link = container_of(cl, struct ipc_link, mbox_cl);
+	struct keembay_ipc_dev *ipc_dev = container_of(link,
+						       struct keembay_ipc_dev,
+						       vpu_link);
+	struct kmb_ipc_pkt *ipc_pkt;
+	struct rx_data *rx_data;
+	struct ipc_chan *chan;
+	u32 entry;
+	int idx;
+
+	entry = *((u32 *)msg);
+
+	/* Get IPC packet. */
+	ipc_pkt = ipc_vpu_to_virt(&ipc_dev->vpu_ipc_mem, entry);
+	if (unlikely(!ipc_pkt)) {
+		dev_warn(dev, "RX: Message out of expected memory range: %x\n",
+			 entry);
+
+		/* Return immediately (cannot mark the IPC packet as free). */
+		return;
+	}
+	if (unlikely(ipc_pkt->src_node != KMB_IPC_NODE_LEON_MSS)) {
+		dev_warn(dev, "RX: Message from unexpected source: %d\n",
+			 ipc_pkt->src_node);
+		goto exit;
+	}
+
+	/* Check destination node. */
+	if (unlikely(ipc_pkt->dst_node != KMB_IPC_NODE_CPU)) {
+		dev_warn(dev, "RX: Message is not for this node\n");
+		goto exit;
+	}
+
+	/* Preliminary channel check. */
+	if (unlikely(ipc_pkt->channel >= KMB_IPC_MAX_CHANNELS)) {
+		dev_warn(dev, "RX: Message for invalid channel\n");
+		goto exit;
+	}
+
+	/* Access internal channel struct (this is protected by an SRCU). */
+	idx = srcu_read_lock(&link->srcu_sp[ipc_pkt->channel]);
+	chan = srcu_dereference(link->channels[ipc_pkt->channel],
+				&link->srcu_sp[ipc_pkt->channel]);
+	if (unlikely(!chan)) {
+		srcu_read_unlock(&link->srcu_sp[ipc_pkt->channel], idx);
+		dev_warn(dev, "RX: Message for closed channel.\n");
+		goto exit;
+	}
+	rx_data = kmalloc(sizeof(*rx_data), GFP_ATOMIC);
+	if (unlikely(!rx_data)) {
+		/* If kmalloc fails, we are forced to discard the message. */
+		srcu_read_unlock(&link->srcu_sp[ipc_pkt->channel], idx);
+		dev_err(dev, "RX: Message dropped: Cannot allocate RX Data.\n");
+		goto exit;
+	}
+	/* Read data info. */
+	rx_data->data_vpu_addr = ipc_pkt->data_addr;
+	rx_data->data_size = ipc_pkt->data_size;
+	/*
+	 * Put data info in rx channel queue.
+	 *
+	 * Note: rx_lock is shared with user context only (and this function is
+	 * run in tasklet context), so spin_lock() is enough.
+	 */
+	spin_lock(&chan->rx_lock);
+	list_add_tail(&rx_data->list, &chan->rx_data_list);
+	spin_unlock(&chan->rx_lock);
+
+	/* Wake up thread waiting on recv(). */
+	wake_up_interruptible(&chan->rx_wait_queue);
+
+	/* Exit SRCU region protecting chan struct. */
+	srcu_read_unlock(&link->srcu_sp[ipc_pkt->channel], idx);
+
+exit:
+	barrier(); /* Ensure IPC packet is fully processed before release. */
+	ipc_pkt->status = KMB_IPC_PKT_FREE;
+}
+
+static void ipc_mbox_tx_done(struct mbox_client *cl, void *mssg, int r)
+{
+	struct tx_data *tx_data = container_of(mssg, struct tx_data, entry);
+	struct ipc_link *link = container_of(cl, struct ipc_link, mbox_cl);
+
+	/* Signal that there is one more free slot in mbox tx queue. */
+	complete(&link->mbox_tx_queue);
+
+	/* Save TX result and notify that IPC TX is completed. */
+	tx_data->retv = r;
+	complete(&tx_data->tx_done);
+}
+
+/**
+ * tx_data_send() - Send a TX data element.
+ * @ipc_dev:	The IPC device to use.
+ * @tx_data:	The TX data element to send.
+ */
+static void tx_data_send(struct keembay_ipc_dev *ipc_dev,
+			 struct tx_data *tx_data)
+{
+	struct device *dev = &ipc_dev->plat_dev->dev;
+	struct ipc_link *link = &ipc_dev->vpu_link;
+	struct kmb_ipc_pkt *ipc_pkt;
+	int rc;
+
+	/* Allocate and set IPC packet. */
+	ipc_pkt = ipc_pkt_tx_alloc(&ipc_dev->ipc_pkt_pool);
+	if (unlikely(!ipc_pkt)) {
+		rc = -ENOMEM;
+		goto error;
+	}
+
+	/* Prepare IPC packet. */
+	ipc_pkt->channel = tx_data->chan_id;
+	ipc_pkt->src_node = KMB_IPC_NODE_CPU;
+	ipc_pkt->dst_node = tx_data->dst_node;
+	ipc_pkt->data_addr = tx_data->vpu_addr;
+	ipc_pkt->data_size = tx_data->size;
+
+	/* Ensure changes to IPC Packet are performed before entry is sent. */
+	wmb();
+
+	/* Initialize entry to ipc_pkt VPU address. */
+	rc = ipc_virt_to_vpu(&ipc_dev->cpu_ipc_mem, ipc_pkt, &tx_data->entry);
+
+	/*
+	 * Check validity of IPC packet VPU address (this error should never
+	 * occur if IPC packet region is defined properly in Device Tree).
+	 */
+	if (unlikely(rc)) {
+		dev_err(dev, "Cannot convert IPC buf vaddr to vpu_addr: %p\n",
+			ipc_pkt);
+		rc = -ENXIO;
+		goto error;
+	}
+	if (unlikely(!IS_ALIGNED(tx_data->entry, KMB_IPC_ALIGNMENT))) {
+		dev_err(dev, "Allocated IPC buf is not 64-byte aligned: %p\n",
+			ipc_pkt);
+		rc = -EFAULT;
+		goto error;
+	}
+
+	/*
+	 * Ensure that the mbox TX queue is not full before passing the packet
+	 * to mbox controller with mbox_send_message(). This will ensure that
+	 * the packet won't be dropped by the mbox framework with the error
+	 * "Try increasing MBOX_TX_QUEUE_LEN".
+	 */
+	rc = wait_for_completion_interruptible(&link->mbox_tx_queue);
+	if (unlikely(rc))
+		goto error;
+
+	mbox_send_message(link->mbox_chan, &tx_data->entry);
+
+	return;
+
+error:
+	/* If an error occurred and a packet was allocated, free it. */
+	if (ipc_pkt)
+		ipc_pkt->status = KMB_IPC_PKT_FREE;
+
+	tx_data->retv = rc;
+	complete(&tx_data->tx_done);
+}
+
+/**
+ * tx_data_dequeue() - Dequeue the next TX data waiting for transfer.
+ * @link:  The link from which we dequeue the TX Data.
+ *
+ * The de-queue policy is round robin between each high speed channel queue and
+ * the one queue for all low speed channels.
+ *
+ * Return: The next TX data waiting to be transferred, or NULL if no TX is
+ *	   pending.
+ */
+static struct tx_data *tx_data_dequeue(struct ipc_link *link)
+{
+	struct tx_data *tx_data;
+	struct tx_queue *queue;
+	int i;
+
+	/*
+	 * TX queues are logically organized in a circular array.
+	 * We go through such an array until we find a non-empty queue.
+	 * We start from where we left since last function invocation.
+	 * If all queues are empty we return NULL.
+	 */
+	for (i = 0; i < ARRAY_SIZE(link->tx_queues); i++) {
+		queue = &link->tx_queues[link->tx_qidx];
+		link->tx_qidx++;
+		if (link->tx_qidx == ARRAY_SIZE(link->tx_queues))
+			link->tx_qidx = 0;
+
+		spin_lock(&queue->lock);
+		tx_data = list_first_entry_or_null(&queue->tx_data_list,
+						   struct tx_data, list);
+		/* If no data in the queue, process the next queue. */
+		if (!tx_data) {
+			spin_unlock(&queue->lock);
+			continue;
+		}
+		/* Otherwise remove rx_data from queue and return. */
+		list_del(&tx_data->list);
+		spin_unlock(&queue->lock);
+
+		return tx_data;
+	}
+
+	return NULL;
+}
+
+/**
+ * tx_data_enqueue() - Enqueue TX data for transfer into the specified link.
+ * @link:	The link the data is enqueued to.
+ * @tx_data:	The TX data to enqueue.
+ */
+static void tx_data_enqueue(struct ipc_link *link, struct tx_data *tx_data)
+{
+	struct tx_queue *queue;
+	int qid;
+
+	/*
+	 * Find the right queue where to put TX data:
+	 * - Each high-speed channel has a dedicated queue, whose index is the
+	 *   same as the channel id (e.g., Channel 1 uses tx_queues[1]).
+	 * - All the general-purpose channels use the same TX queue, which is
+	 *   the last element in the tx_queues array.
+	 *
+	 * Note: tx_queues[] has KMB_IPC_NUM_HS_CHANNELS+1 elements)
+	 */
+	qid = tx_data->chan_id < ARRAY_SIZE(link->tx_queues) ?
+	      tx_data->chan_id : (ARRAY_SIZE(link->tx_queues) - 1);
+
+	queue = &link->tx_queues[qid];
+
+	/*
+	 * Lock shared between user contexts (callers of ipc_send()) and tx
+	 * thread; so spin_lock() is enough.
+	 */
+	spin_lock(&queue->lock);
+	list_add_tail(&tx_data->list, &queue->tx_data_list);
+	spin_unlock(&queue->lock);
+}
+
+/**
+ * tx_data_remove() - Remove TX data element from specified link.
+ * @link:	The link the data is currently enqueued to.
+ * @tx_data:	The TX data element to be removed.
+ *
+ * This function is called by the main send function, when the send is
+ * interrupted or has timed out.
+ */
+static void tx_data_remove(struct ipc_link *link, struct tx_data *tx_data)
+{
+	struct tx_queue *queue;
+	int qid;
+
+	/*
+	 * Find the TX queue where TX data is currently located:
+	 * - Each high-speed channel has a dedicated queue, whose index is the
+	 *   same as the channel id (e.g., Channel 1 uses tx_queues[1]).
+	 * - All the general-purpose channels use the same TX queue, which is
+	 *   the last element in the tx_queues array.
+	 *
+	 * Note: tx_queues[] has KMB_IPC_NUM_HS_CHANNELS+1 elements)
+	 */
+	qid = tx_data->chan_id < ARRAY_SIZE(link->tx_queues) ?
+	      tx_data->chan_id : (ARRAY_SIZE(link->tx_queues) - 1);
+
+	queue = &link->tx_queues[qid];
+
+	/*
+	 * Lock shared between user contexts (callers of ipc_send()) and tx
+	 * thread; so spin_lock() is enough.
+	 */
+	spin_lock(&queue->lock);
+	list_del(&tx_data->list);
+	spin_unlock(&queue->lock);
+}
+
+/**
+ * tx_thread_fn() - The function run by the TX thread.
+ * @ptr: A pointer to the keembay_ipc_dev struct associated with the thread.
+ *
+ * This thread continuously dequeues and send TX data elements. The TX
+ * semaphore is used to pause the loop when all the pending TX data elements
+ * have been transmitted (the send function 'ups' the semaphore every time a
+ * new TX data element is enqueued).
+ */
+static int tx_thread_fn(void *ptr)
+{
+	struct keembay_ipc_dev *ipc_dev = ptr;
+	struct ipc_link *link = &ipc_dev->vpu_link;
+	struct tx_data *tx_data;
+	int rc;
+
+	while (1) {
+		rc = wait_for_completion_interruptible(&link->tx_queued);
+		if (rc || link->tx_stopping)
+			break;
+		tx_data = tx_data_dequeue(link);
+		/*
+		 * We can get a null tx_data if tx_data_remove() has been
+		 * called. Just ignore it and continue.
+		 */
+		if (!tx_data)
+			continue;
+		tx_data_send(ipc_dev, tx_data);
+	}
+
+	/* Wait until kthread_stop() is called. */
+	set_current_state(TASK_INTERRUPTIBLE);
+	while (!kthread_should_stop()) {
+		schedule();
+		set_current_state(TASK_INTERRUPTIBLE);
+	}
+	__set_current_state(TASK_RUNNING);
+
+	return rc;
+}
+
+/* Internal send. */
+static int __ipc_send(struct keembay_ipc_dev *ipc_dev, u8 dst_node,
+		      u16 chan_id, u32 vpu_addr, size_t size)
+{
+	struct ipc_link *link = &ipc_dev->vpu_link;
+	struct tx_data *tx_data;
+	int rc;
+
+	/* Allocate and init TX data. */
+	tx_data = kmalloc(sizeof(*tx_data), GFP_KERNEL);
+	if (!tx_data)
+		return -ENOMEM;
+	tx_data->dst_node = dst_node;
+	tx_data->chan_id = chan_id;
+	tx_data->vpu_addr = vpu_addr;
+	tx_data->size = size;
+	tx_data->retv = 1;
+	INIT_LIST_HEAD(&tx_data->list);
+	init_completion(&tx_data->tx_done);
+
+	/* Add tx_data to tx queues. */
+	tx_data_enqueue(link, tx_data);
+
+	/* Signal that we have a new pending TX. */
+	complete(&link->tx_queued);
+
+	/* Wait until data is transmitted. */
+	rc = wait_for_completion_interruptible(&tx_data->tx_done);
+	if (unlikely(rc)) {
+		tx_data_remove(link, tx_data);
+		goto exit;
+	}
+	rc = tx_data->retv;
+
+exit:
+	kfree(tx_data);
+	return rc;
+}
+
+/*
+ * Driver allocation.
+ */
+
+/* Device tree driver match. */
+static const struct of_device_id kmb_ipc_of_match[] = {
+	{
+		.compatible = "intel,keembay-ipc",
+	},
+	{}
+};
+
+/* The IPC driver is a platform device. */
+static struct platform_driver kmb_ipc_driver = {
+	.probe  = kmb_ipc_probe,
+	.remove = kmb_ipc_remove,
+	.driver = {
+		.name = DRV_NAME,
+		.of_match_table = kmb_ipc_of_match,
+	},
+};
+
+module_platform_driver(kmb_ipc_driver);
+
+/*
+ * Perform basic validity check on common API arguments.
+ *
+ * Verify that the specified device is a Keem Bay IPC device and that node ID
+ * and the channel ID are within the allowed ranges.
+ */
+static int validate_api_args(struct device *dev, u8 node_id, u16 chan_id)
+{
+	if (!dev || dev->driver != &kmb_ipc_driver.driver)
+		return -EINVAL;
+	if (node_id != KMB_IPC_NODE_LEON_MSS) {
+		dev_warn(dev, "Invalid Link ID\n");
+		return -EINVAL;
+	}
+	if (chan_id >= KMB_IPC_MAX_CHANNELS) {
+		dev_warn(dev, "Invalid Channel ID\n");
+		return -EINVAL;
+	}
+	return 0;
+}
+
+/*
+ * IPC Kernel API.
+ */
+
+/**
+ * intel_keembay_ipc_open_channel() - Open an IPC channel.
+ * @dev:	The IPC device to use.
+ * @node_id:	The node ID of the remote node (used to identify the link the
+ *		channel must be added to). KMB_IPC_NODE_LEON_MSS is the only
+ *		allowed value for now.
+ * @chan_id:	The ID of the channel to be opened.
+ *
+ * Return:	0 on success, negative error code otherwise.
+ */
+int intel_keembay_ipc_open_channel(struct device *dev, u8 node_id, u16 chan_id)
+{
+	struct ipc_chan *chan, *cur_chan;
+	struct keembay_ipc_dev *ipc_dev;
+	struct ipc_link *link;
+	int rc;
+
+	rc = validate_api_args(dev, node_id, chan_id);
+	if (rc)
+		return rc;
+
+	ipc_dev = dev_get_drvdata(dev);
+	link = &ipc_dev->vpu_link;
+
+	/* Create channel before getting lock. */
+	chan = kzalloc(sizeof(*chan), GFP_KERNEL);
+	if (!chan)
+		return -ENOMEM;
+
+	INIT_LIST_HEAD(&chan->rx_data_list);
+	spin_lock_init(&chan->rx_lock);
+	init_waitqueue_head(&chan->rx_wait_queue);
+
+	/* Add channel to the channel array (if not already present). */
+	spin_lock(&link->chan_lock);
+	cur_chan = rcu_dereference_protected(link->channels[chan_id],
+					     lockdep_is_held(&link->chan_lock));
+	if (cur_chan) {
+		spin_unlock(&link->chan_lock);
+		kfree(chan);
+		return -EEXIST;
+	}
+	rcu_assign_pointer(link->channels[chan_id], chan);
+	spin_unlock(&link->chan_lock);
+
+	return 0;
+}
+EXPORT_SYMBOL(intel_keembay_ipc_open_channel);
+
+/**
+ * intel_keembay_ipc_close_channel() - Close an IPC channel.
+ * @dev:	The IPC device to use.
+ * @node_id:	The node ID of the remote node (used to identify the link the
+ *		channel must be added to). KMB_IPC_NODE_LEON_MSS is the only
+ *		allowed value for now.
+ * @chan_id:	The ID of the channel to be closed.
+ *
+ * Return:	0 on success, negative error code otherwise.
+ */
+int intel_keembay_ipc_close_channel(struct device *dev, u8 node_id, u16 chan_id)
+{
+	struct keembay_ipc_dev *ipc_dev;
+	struct ipc_link *link;
+	int rc;
+
+	rc = validate_api_args(dev, node_id, chan_id);
+	if (rc)
+		return rc;
+
+	ipc_dev = dev_get_drvdata(dev);
+	link = &ipc_dev->vpu_link;
+
+	rc = channel_close(link, chan_id);
+	if (!rc)
+		dev_info(dev, "Channel was already closed\n");
+
+	return 0;
+}
+EXPORT_SYMBOL(intel_keembay_ipc_close_channel);
+
+/**
+ * intel_keembay_ipc_send() - Send data via IPC.
+ * @dev:	The IPC device to use.
+ * @node_id:	The node ID of the remote node (used to identify the link the
+ *		channel must be added to). KMB_IPC_NODE_LEON_MSS is the only
+ *		allowed value for now.
+ * @chan_id:	The IPC channel to be used to send the message.
+ * @vpu_addr:	The VPU address of the data to be transferred.
+ * @size:	The size of the data to be transferred.
+ *
+ * Return:	0 on success, negative error code otherwise.
+ */
+int intel_keembay_ipc_send(struct device *dev, u8 node_id, u16 chan_id,
+			   u32 vpu_addr, size_t size)
+{
+	struct keembay_ipc_dev *ipc_dev;
+	struct ipc_link *link;
+	struct ipc_chan *chan;
+	int idx, rc;
+
+	rc = validate_api_args(dev, node_id, chan_id);
+	if (rc)
+		return rc;
+
+	ipc_dev = dev_get_drvdata(dev);
+	link = &ipc_dev->vpu_link;
+	/*
+	 * Start Sleepable RCU critical section (this prevents close() from
+	 * destroying the channels struct while we are sending data)
+	 */
+	idx = srcu_read_lock(&link->srcu_sp[chan_id]);
+
+	/* Get channel. */
+	chan = srcu_dereference(link->channels[chan_id],
+				&link->srcu_sp[chan_id]);
+	if (unlikely(!chan)) {
+		/* The channel is closed. */
+		rc = -ENOENT;
+		goto exit;
+	}
+
+	rc = __ipc_send(ipc_dev, node_id, chan_id, vpu_addr, size);
+
+exit:
+	/* End sleepable RCU critical section. */
+	srcu_read_unlock(&link->srcu_sp[chan_id], idx);
+	return rc;
+}
+EXPORT_SYMBOL(intel_keembay_ipc_send);
+
+/**
+ * intel_keembay_ipc_recv() - Read data via IPC
+ * @dev:	The IPC device to use.
+ * @node_id:	The node ID of the remote node (used to identify the link the
+ *		channel must be added to). KMB_IPC_NODE_LEON_MSS is the only
+ *		allowed value for now.
+ * @chan_id:	The IPC channel to read from.
+ * @vpu_addr:	[out] The VPU address of the received data.
+ * @size:	[out] Where to store the size of the received data.
+ * @timeout:	How long (in ms) the function will block waiting for an IPC
+ *		message; if UINT32_MAX it will block indefinitely; if 0 it
+ *		will not block.
+ *
+ * Return:	0 on success, negative error code otherwise
+ */
+int intel_keembay_ipc_recv(struct device *dev, u8 node_id, u16 chan_id,
+			   u32 *vpu_addr, size_t *size, u32 timeout)
+{
+	struct keembay_ipc_dev *ipc_dev;
+	struct rx_data *rx_entry;
+	struct ipc_link *link;
+	struct ipc_chan *chan;
+	int idx, rc;
+
+	rc = validate_api_args(dev, node_id, chan_id);
+	if (rc)
+		return rc;
+
+	if (!vpu_addr || !size)
+		return -EINVAL;
+
+	ipc_dev = dev_get_drvdata(dev);
+	link = &ipc_dev->vpu_link;
+
+	/*
+	 * Start Sleepable RCU critical section (this prevents close() from
+	 * destroying the channels struct while we are using it)
+	 */
+	idx = srcu_read_lock(&link->srcu_sp[chan_id]);
+
+	/* Get channel. */
+	chan = srcu_dereference(link->channels[chan_id],
+				&link->srcu_sp[chan_id]);
+	if (unlikely(!chan)) {
+		rc = -ENOENT;
+		goto err;
+	}
+	/*
+	 * Get the lock protecting rx_data_list; the lock will be released by
+	 * wait_event_*_lock_irq() before going to sleep and automatically
+	 * reacquired after wake up.
+	 *
+	 * NOTE: lock_irq() is needed because rx_lock is also used by the RX
+	 * tasklet; also lock_bh() is not used because there is no
+	 * wait_event_interruptible_lock_bh().
+	 */
+	spin_lock_irq(&chan->rx_lock);
+	/*
+	 * Wait for RX data.
+	 *
+	 * Note: wait_event_interruptible_lock_irq_timeout() has different
+	 * return values than wait_event_interruptible_lock_irq().
+	 *
+	 * The following if/then branch ensures that return values are
+	 * consistent for the both cases, that is:
+	 * - rc == 0 only if the wait was successfully (i.e., we were notified
+	 *   of a message or of a channel closure)
+	 * - rc < 0 if an error occurred (we got interrupted or the timeout
+	 *   expired).
+	 */
+	if (timeout == U32_MAX) {
+		rc = wait_event_interruptible_lock_irq(chan->rx_wait_queue,
+						       !list_empty(&chan->rx_data_list) ||
+						       chan->closing,
+						       chan->rx_lock);
+	} else {
+		rc = wait_event_interruptible_lock_irq_timeout(chan->rx_wait_queue,
+							       !list_empty(&chan->rx_data_list) ||
+							       chan->closing,
+							       chan->rx_lock,
+							       msecs_to_jiffies(timeout));
+		if (!rc)
+			rc = -ETIME;
+		if (rc > 0)
+			rc = 0;
+	}
+
+	/* Check if the channel was closed while waiting. */
+	if (chan->closing)
+		rc = -EPIPE;
+	if (rc) {
+		spin_unlock_irq(&chan->rx_lock);
+		goto err;
+	}
+
+	/* Extract RX entry. */
+	rx_entry = list_first_entry(&chan->rx_data_list, struct rx_data, list);
+	list_del(&rx_entry->list);
+	spin_unlock_irq(&chan->rx_lock);
+
+	/* Set output parameters. */
+	*vpu_addr =  rx_entry->data_vpu_addr;
+	*size = rx_entry->data_size;
+
+	/* Free RX entry. */
+	kfree(rx_entry);
+
+err:
+	/* End sleepable RCU critical section. */
+	srcu_read_unlock(&link->srcu_sp[chan_id], idx);
+	return rc;
+}
+EXPORT_SYMBOL(intel_keembay_ipc_recv);
+
+MODULE_DESCRIPTION("Keem Bay IPC Driver");
+MODULE_AUTHOR("Daniele Alessandrelli <daniele.alessandrelli@intel.com>");
+MODULE_AUTHOR("Paul Murphy <paul.j.murphy@intel.com>");
+MODULE_LICENSE("GPL");
diff --git a/include/linux/soc/intel/keembay-ipc.h b/include/linux/soc/intel/keembay-ipc.h
new file mode 100644
index 000000000000..ac7748d1595f
--- /dev/null
+++ b/include/linux/soc/intel/keembay-ipc.h
@@ -0,0 +1,30 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Keem Bay IPC Linux Kernel API
+ *
+ * Copyright (C) 2018-2020 Intel Corporation
+ */
+
+#ifndef __KEEMBAY_IPC_H
+#define __KEEMBAY_IPC_H
+
+#include <linux/types.h>
+
+/* The possible node IDs. */
+enum {
+	KMB_IPC_NODE_CPU = 0,
+	KMB_IPC_NODE_LEON_MSS,
+};
+
+int intel_keembay_ipc_open_channel(struct device *dev, u8 node_id, u16 chan_id);
+
+int intel_keembay_ipc_close_channel(struct device *dev, u8 node_id,
+				    u16 chan_id);
+
+int intel_keembay_ipc_send(struct device *dev, u8 node_id, u16 chan_id,
+			   u32 vpu_addr, size_t size);
+
+int intel_keembay_ipc_recv(struct device *dev, u8 node_id, u16 chan_id,
+			   u32 *vpu_addr, size_t *size, u32 timeout);
+
+#endif /* __KEEMBAY_IPC_H */
-- 
2.17.1


  parent reply index

Thread overview: 57+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-01-08 21:25 [PATCH v2 00/34] Intel Vision Processing base enabling mgross
2021-01-08 21:25 ` [PATCH v2 01/34] Add Vision Processing Unit (VPU) documentation mgross
2021-01-08 21:25 ` [PATCH v2 02/34] dt-bindings: mailbox: Add Intel VPU IPC mailbox bindings mgross
2021-01-08 21:25 ` [PATCH v2 03/34] mailbox: vpu-ipc-mailbox: Add support for Intel VPU IPC mailbox mgross
2021-01-08 21:25 ` [PATCH v2 04/34] dt-bindings: Add bindings for Keem Bay IPC driver mgross
2021-01-08 21:25 ` mgross [this message]
2021-01-08 21:25 ` [PATCH v2 06/34] dt-bindings: Add bindings for Keem Bay VPU " mgross
2021-01-10 17:18   ` Rob Herring
2021-01-11 19:24   ` Rob Herring
2021-01-19 14:32     ` Alessandrelli, Daniele
2021-01-08 21:25 ` [PATCH v2 07/34] keembay-vpu-ipc: Add Keem Bay VPU IPC module mgross
2021-01-08 21:25 ` [PATCH v2 08/34] misc: xlink-pcie: Add documentation for XLink PCIe driver mgross
2021-01-19 19:36   ` Randy Dunlap
2021-01-24 18:27     ` Thokala, Srikanth
2021-01-08 21:25 ` [PATCH v2 09/34] misc: xlink-pcie: lh: Add PCIe EPF driver for Local Host mgross
2021-01-20 17:57   ` Greg KH
2021-01-24 11:48     ` Thokala, Srikanth
2021-01-24 11:56       ` Greg KH
2021-01-24 18:18         ` Thokala, Srikanth
2021-01-08 21:25 ` [PATCH v2 10/34] misc: xlink-pcie: lh: Add PCIe EP DMA functionality mgross
2021-01-08 21:25 ` [PATCH v2 11/34] misc: xlink-pcie: lh: Add core communication logic mgross
2021-01-08 21:25 ` [PATCH v2 12/34] misc: xlink-pcie: lh: Prepare changes for adding remote host driver mgross
2021-01-08 21:25 ` [PATCH v2 13/34] misc: xlink-pcie: rh: Add PCIe EP driver for Remote Host mgross
2021-01-08 21:25 ` [PATCH v2 14/34] misc: xlink-pcie: rh: Add core communication logic mgross
2021-01-08 21:25 ` [PATCH v2 15/34] misc: xlink-pcie: Add XLink API interface mgross
2021-01-20 17:59   ` Greg KH
2021-01-21 23:20     ` mark gross
2021-01-24 11:46     ` Thokala, Srikanth
2021-01-08 21:25 ` [PATCH v2 16/34] misc: xlink-pcie: Add asynchronous event notification support for XLink mgross
2021-01-08 21:25 ` [PATCH v2 17/34] xlink-ipc: Add xlink ipc device tree bindings mgross
2021-01-10 17:18   ` Rob Herring
2021-01-08 21:25 ` [PATCH v2 18/34] xlink-ipc: Add xlink ipc driver mgross
2021-01-08 21:25 ` [PATCH v2 19/34] xlink-core: Add xlink core device tree bindings mgross
2021-01-10 17:18   ` Rob Herring
2021-01-11 19:27   ` Rob Herring
2021-01-08 21:25 ` [PATCH v2 20/34] xlink-core: Add xlink core driver xLink mgross
2021-01-19 19:58   ` Randy Dunlap
2021-01-08 21:25 ` [PATCH v2 21/34] xlink-core: Enable xlink protocol over pcie mgross
2021-01-08 21:25 ` [PATCH v2 22/34] xlink-core: Enable VPU IP management and runtime control mgross
2021-01-08 21:25 ` [PATCH v2 23/34] xlink-core: add async channel and events mgross
2021-01-08 21:25 ` [PATCH v2 24/34] dt-bindings: misc: Add Keem Bay vpumgr mgross
2021-01-08 21:25 ` [PATCH v2 25/34] misc: Add Keem Bay VPU manager mgross
2021-01-08 21:25 ` [PATCH v2 26/34] dt-bindings: misc: intel_tsens: Add tsens thermal bindings documentation mgross
2021-01-08 21:25 ` [PATCH v2 27/34] misc: Tsens ARM host thermal driver mgross
2021-01-08 21:25 ` [PATCH v2 28/34] misc: Intel tsens IA host driver mgross
2021-01-08 21:25 ` [PATCH v2 29/34] Intel tsens i2c slave driver mgross
2021-01-12  7:15   ` Randy Dunlap
2021-01-25 23:39     ` mark gross
2021-01-26  7:45       ` Arnd Bergmann
2021-01-26 14:56         ` Gross, Mark
2021-01-27  4:45         ` C, Udhayakumar
2021-01-27  4:44       ` C, Udhayakumar
2021-01-08 21:25 ` [PATCH v2 30/34] misc:intel_tsens: Intel Keem Bay tsens driver mgross
2021-01-08 21:25 ` [PATCH v2 31/34] Intel Keem Bay XLink SMBus driver mgross
2021-01-08 21:25 ` [PATCH v2 32/34] dt-bindings: misc: hddl_dev: Add hddl device management documentation mgross
2021-01-08 21:25 ` [PATCH v2 33/34] misc: Hddl device management for local host mgross
2021-01-08 21:26 ` [PATCH v2 34/34] misc: HDDL device management for IA host mgross

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20210108212600.36850-6-mgross@linux.intel.com \
    --to=mgross@linux.intel.com \
    --cc=arnd@arndb.de \
    --cc=bp@suse.de \
    --cc=corbet@lwn.net \
    --cc=damien.lemoal@wdc.com \
    --cc=daniele.alessandrelli@intel.com \
    --cc=dragan.cvetic@xilinx.com \
    --cc=gregkh@linuxfoundation.org \
    --cc=jassisinghbrar@gmail.com \
    --cc=leonard.crestez@nxp.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=markgross@kernel.org \
    --cc=palmerdabbelt@google.com \
    --cc=paul.walmsley@sifive.com \
    --cc=peng.fan@nxp.com \
    --cc=robh+dt@kernel.org \
    --cc=shawnguo@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

LKML Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/lkml/0 lkml/git/0.git
	git clone --mirror https://lore.kernel.org/lkml/1 lkml/git/1.git
	git clone --mirror https://lore.kernel.org/lkml/2 lkml/git/2.git
	git clone --mirror https://lore.kernel.org/lkml/3 lkml/git/3.git
	git clone --mirror https://lore.kernel.org/lkml/4 lkml/git/4.git
	git clone --mirror https://lore.kernel.org/lkml/5 lkml/git/5.git
	git clone --mirror https://lore.kernel.org/lkml/6 lkml/git/6.git
	git clone --mirror https://lore.kernel.org/lkml/7 lkml/git/7.git
	git clone --mirror https://lore.kernel.org/lkml/8 lkml/git/8.git
	git clone --mirror https://lore.kernel.org/lkml/9 lkml/git/9.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 lkml lkml/ https://lore.kernel.org/lkml \
		linux-kernel@vger.kernel.org
	public-inbox-index lkml

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-kernel


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git