Merge tag 'tegra-for-4.10-firmware' of git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux into next/drivers

firmware: Add Tegra IVC and BPMP support

IVC is an inter-processor communication protocol that uses shared memory
to exchange data between processors. The BPMP driver makes use of this
to communicate with the Boot and Power Management Processor (BPMP) and
uses an additional hardware synchronization primitive from the HSP block
to signal availability of new data (doorbell).

Firmware running on the BPMP implements a number of services such as the
control of clocks and resets within the system, or the ability to ungate
or gate power partitions.

* tag 'tegra-for-4.10-firmware' of git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux:
  dt-bindings: firmware: Allow child nodes inside the Tegra BPMP
  dt-bindings: Add power domains to Tegra BPMP firmware
  firmware: tegra: Add BPMP support
  firmware: tegra: Add IVC library
  dt-bindings: firmware: Add bindings for Tegra BPMP

Signed-off-by: Olof Johansson <olof@lixom.net>

6 files changed, 1592 insertions, 0 deletions

diff --git a/drivers/firmware/Kconfig b/drivers/firmware/Kconfig
index bca172d..9968de04 100644
--- a/drivers/firmware/Kconfig
+++ b/drivers/firmware/Kconfig
@@ -210,5 +210,6 @@ source "drivers/firmware/broadcom/Kconfig"
 source "drivers/firmware/google/Kconfig"
 source "drivers/firmware/efi/Kconfig"
 source "drivers/firmware/meson/Kconfig"
+source "drivers/firmware/tegra/Kconfig"
 
 endmenu
diff --git a/drivers/firmware/Makefile b/drivers/firmware/Makefile
index 898ac41..2afe75c5 100644
--- a/drivers/firmware/Makefile
+++ b/drivers/firmware/Makefile
@@ -26,3 +26,4 @@ obj-y				+= meson/
 obj-$(CONFIG_GOOGLE_FIRMWARE)	+= google/
 obj-$(CONFIG_EFI)		+= efi/
 obj-$(CONFIG_UEFI_CPER)		+= efi/
+obj-y				+= tegra/
diff --git a/drivers/firmware/tegra/Kconfig b/drivers/firmware/tegra/Kconfig
new file mode 100644
index 0000000..ff2730d5
--- /dev/null
+++ b/drivers/firmware/tegra/Kconfig
@@ -0,0 +1,25 @@
+menu "Tegra firmware driver"
+
+config TEGRA_IVC
+	bool "Tegra IVC protocol"
+	depends on ARCH_TEGRA
+	help
+	  IVC (Inter-VM Communication) protocol is part of the IPC
+	  (Inter Processor Communication) framework on Tegra. It maintains the
+	  data and the different commuication channels in SysRAM or RAM and
+	  keeps the content is synchronization between host CPU and remote
+	  processors.
+
+config TEGRA_BPMP
+	bool "Tegra BPMP driver"
+	depends on ARCH_TEGRA && TEGRA_HSP_MBOX && TEGRA_IVC
+	help
+	  BPMP (Boot and Power Management Processor) is designed to off-loading
+	  the PM functions which include clock/DVFS/thermal/power from the CPU.
+	  It needs HSP as the HW synchronization and notification module and
+	  IVC module as the message communication protocol.
+
+	  This driver manages the IPC interface between host CPU and the
+	  firmware running on BPMP.
+
+endmenu
diff --git a/drivers/firmware/tegra/Makefile b/drivers/firmware/tegra/Makefile
new file mode 100644
index 0000000..e34a2f7
--- /dev/null
+++ b/drivers/firmware/tegra/Makefile
@@ -0,0 +1,2 @@
+obj-$(CONFIG_TEGRA_BPMP)	+= bpmp.o
+obj-$(CONFIG_TEGRA_IVC)		+= ivc.o
diff --git a/drivers/firmware/tegra/bpmp.c b/drivers/firmware/tegra/bpmp.c
new file mode 100644
index 0000000..4ff02d3
--- /dev/null
+++ b/drivers/firmware/tegra/bpmp.c
@@ -0,0 +1,868 @@
+/*
+ * Copyright (c) 2016, NVIDIA CORPORATION.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope 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 for
+ * more details.
+ */
+
+#include <linux/clk/tegra.h>
+#include <linux/genalloc.h>
+#include <linux/mailbox_client.h>
+#include <linux/of.h>
+#include <linux/of_address.h>
+#include <linux/of_device.h>
+#include <linux/platform_device.h>
+#include <linux/semaphore.h>
+
+#include <soc/tegra/bpmp.h>
+#include <soc/tegra/bpmp-abi.h>
+#include <soc/tegra/ivc.h>
+
+#define MSG_ACK		BIT(0)
+#define MSG_RING	BIT(1)
+
+static inline struct tegra_bpmp *
+mbox_client_to_bpmp(struct mbox_client *client)
+{
+	return container_of(client, struct tegra_bpmp, mbox.client);
+}
+
+struct tegra_bpmp *tegra_bpmp_get(struct device *dev)
+{
+	struct platform_device *pdev;
+	struct tegra_bpmp *bpmp;
+	struct device_node *np;
+
+	np = of_parse_phandle(dev->of_node, "nvidia,bpmp", 0);
+	if (!np)
+		return ERR_PTR(-ENOENT);
+
+	pdev = of_find_device_by_node(np);
+	if (!pdev) {
+		bpmp = ERR_PTR(-ENODEV);
+		goto put;
+	}
+
+	bpmp = platform_get_drvdata(pdev);
+	if (!bpmp) {
+		bpmp = ERR_PTR(-EPROBE_DEFER);
+		put_device(&pdev->dev);
+		goto put;
+	}
+
+put:
+	of_node_put(np);
+	return bpmp;
+}
+EXPORT_SYMBOL_GPL(tegra_bpmp_get);
+
+void tegra_bpmp_put(struct tegra_bpmp *bpmp)
+{
+	if (bpmp)
+		put_device(bpmp->dev);
+}
+EXPORT_SYMBOL_GPL(tegra_bpmp_put);
+
+static int tegra_bpmp_channel_get_index(struct tegra_bpmp_channel *channel)
+{
+	return channel - channel->bpmp->channels;
+}
+
+static int
+tegra_bpmp_channel_get_thread_index(struct tegra_bpmp_channel *channel)
+{
+	struct tegra_bpmp *bpmp = channel->bpmp;
+	unsigned int offset, count;
+	int index;
+
+	offset = bpmp->soc->channels.thread.offset;
+	count = bpmp->soc->channels.thread.count;
+
+	index = tegra_bpmp_channel_get_index(channel);
+	if (index < 0)
+		return index;
+
+	if (index < offset || index >= offset + count)
+		return -EINVAL;
+
+	return index - offset;
+}
+
+static struct tegra_bpmp_channel *
+tegra_bpmp_channel_get_thread(struct tegra_bpmp *bpmp, unsigned int index)
+{
+	unsigned int offset = bpmp->soc->channels.thread.offset;
+	unsigned int count = bpmp->soc->channels.thread.count;
+
+	if (index >= count)
+		return NULL;
+
+	return &bpmp->channels[offset + index];
+}
+
+static struct tegra_bpmp_channel *
+tegra_bpmp_channel_get_tx(struct tegra_bpmp *bpmp)
+{
+	unsigned int offset = bpmp->soc->channels.cpu_tx.offset;
+
+	return &bpmp->channels[offset + smp_processor_id()];
+}
+
+static struct tegra_bpmp_channel *
+tegra_bpmp_channel_get_rx(struct tegra_bpmp *bpmp)
+{
+	unsigned int offset = bpmp->soc->channels.cpu_rx.offset;
+
+	return &bpmp->channels[offset];
+}
+
+static bool tegra_bpmp_message_valid(const struct tegra_bpmp_message *msg)
+{
+	return (msg->tx.size <= MSG_DATA_MIN_SZ) &&
+	       (msg->rx.size <= MSG_DATA_MIN_SZ) &&
+	       (msg->tx.size == 0 || msg->tx.data) &&
+	       (msg->rx.size == 0 || msg->rx.data);
+}
+
+static bool tegra_bpmp_master_acked(struct tegra_bpmp_channel *channel)
+{
+	void *frame;
+
+	frame = tegra_ivc_read_get_next_frame(channel->ivc);
+	if (IS_ERR(frame)) {
+		channel->ib = NULL;
+		return false;
+	}
+
+	channel->ib = frame;
+
+	return true;
+}
+
+static int tegra_bpmp_wait_ack(struct tegra_bpmp_channel *channel)
+{
+	unsigned long timeout = channel->bpmp->soc->channels.cpu_tx.timeout;
+	ktime_t end;
+
+	end = ktime_add_us(ktime_get(), timeout);
+
+	do {
+		if (tegra_bpmp_master_acked(channel))
+			return 0;
+	} while (ktime_before(ktime_get(), end));
+
+	return -ETIMEDOUT;
+}
+
+static bool tegra_bpmp_master_free(struct tegra_bpmp_channel *channel)
+{
+	void *frame;
+
+	frame = tegra_ivc_write_get_next_frame(channel->ivc);
+	if (IS_ERR(frame)) {
+		channel->ob = NULL;
+		return false;
+	}
+
+	channel->ob = frame;
+
+	return true;
+}
+
+static int tegra_bpmp_wait_master_free(struct tegra_bpmp_channel *channel)
+{
+	unsigned long timeout = channel->bpmp->soc->channels.cpu_tx.timeout;
+	ktime_t start, now;
+
+	start = ns_to_ktime(local_clock());
+
+	do {
+		if (tegra_bpmp_master_free(channel))
+			return 0;
+
+		now = ns_to_ktime(local_clock());
+	} while (ktime_us_delta(now, start) < timeout);
+
+	return -ETIMEDOUT;
+}
+
+static ssize_t __tegra_bpmp_channel_read(struct tegra_bpmp_channel *channel,
+					 void *data, size_t size)
+{
+	if (data && size > 0)
+		memcpy(data, channel->ib->data, size);
+
+	return tegra_ivc_read_advance(channel->ivc);
+}
+
+static ssize_t tegra_bpmp_channel_read(struct tegra_bpmp_channel *channel,
+				       void *data, size_t size)
+{
+	struct tegra_bpmp *bpmp = channel->bpmp;
+	unsigned long flags;
+	ssize_t err;
+	int index;
+
+	index = tegra_bpmp_channel_get_thread_index(channel);
+	if (index < 0)
+		return index;
+
+	spin_lock_irqsave(&bpmp->lock, flags);
+	err = __tegra_bpmp_channel_read(channel, data, size);
+	clear_bit(index, bpmp->threaded.allocated);
+	spin_unlock_irqrestore(&bpmp->lock, flags);
+
+	up(&bpmp->threaded.lock);
+
+	return err;
+}
+
+static ssize_t __tegra_bpmp_channel_write(struct tegra_bpmp_channel *channel,
+					  unsigned int mrq, unsigned long flags,
+					  const void *data, size_t size)
+{
+	channel->ob->code = mrq;
+	channel->ob->flags = flags;
+
+	if (data && size > 0)
+		memcpy(channel->ob->data, data, size);
+
+	return tegra_ivc_write_advance(channel->ivc);
+}
+
+static struct tegra_bpmp_channel *
+tegra_bpmp_write_threaded(struct tegra_bpmp *bpmp, unsigned int mrq,
+			  const void *data, size_t size)
+{
+	unsigned long timeout = bpmp->soc->channels.thread.timeout;
+	unsigned int count = bpmp->soc->channels.thread.count;
+	struct tegra_bpmp_channel *channel;
+	unsigned long flags;
+	unsigned int index;
+	int err;
+
+	err = down_timeout(&bpmp->threaded.lock, usecs_to_jiffies(timeout));
+	if (err < 0)
+		return ERR_PTR(err);
+
+	spin_lock_irqsave(&bpmp->lock, flags);
+
+	index = find_first_zero_bit(bpmp->threaded.allocated, count);
+	if (index == count) {
+		channel = ERR_PTR(-EBUSY);
+		goto unlock;
+	}
+
+	channel = tegra_bpmp_channel_get_thread(bpmp, index);
+	if (!channel) {
+		channel = ERR_PTR(-EINVAL);
+		goto unlock;
+	}
+
+	if (!tegra_bpmp_master_free(channel)) {
+		channel = ERR_PTR(-EBUSY);
+		goto unlock;
+	}
+
+	set_bit(index, bpmp->threaded.allocated);
+
+	err = __tegra_bpmp_channel_write(channel, mrq, MSG_ACK | MSG_RING,
+					 data, size);
+	if (err < 0) {
+		clear_bit(index, bpmp->threaded.allocated);
+		goto unlock;
+	}
+
+	set_bit(index, bpmp->threaded.busy);
+
+unlock:
+	spin_unlock_irqrestore(&bpmp->lock, flags);
+	return channel;
+}
+
+static ssize_t tegra_bpmp_channel_write(struct tegra_bpmp_channel *channel,
+					unsigned int mrq, unsigned long flags,
+					const void *data, size_t size)
+{
+	int err;
+
+	err = tegra_bpmp_wait_master_free(channel);
+	if (err < 0)
+		return err;
+
+	return __tegra_bpmp_channel_write(channel, mrq, flags, data, size);
+}
+
+int tegra_bpmp_transfer_atomic(struct tegra_bpmp *bpmp,
+			       struct tegra_bpmp_message *msg)
+{
+	struct tegra_bpmp_channel *channel;
+	int err;
+
+	if (WARN_ON(!irqs_disabled()))
+		return -EPERM;
+
+	if (!tegra_bpmp_message_valid(msg))
+		return -EINVAL;
+
+	channel = tegra_bpmp_channel_get_tx(bpmp);
+
+	err = tegra_bpmp_channel_write(channel, msg->mrq, MSG_ACK,
+				       msg->tx.data, msg->tx.size);
+	if (err < 0)
+		return err;
+
+	err = mbox_send_message(bpmp->mbox.channel, NULL);
+	if (err < 0)
+		return err;
+
+	mbox_client_txdone(bpmp->mbox.channel, 0);
+
+	err = tegra_bpmp_wait_ack(channel);
+	if (err < 0)
+		return err;
+
+	return __tegra_bpmp_channel_read(channel, msg->rx.data, msg->rx.size);
+}
+EXPORT_SYMBOL_GPL(tegra_bpmp_transfer_atomic);
+
+int tegra_bpmp_transfer(struct tegra_bpmp *bpmp,
+			struct tegra_bpmp_message *msg)
+{
+	struct tegra_bpmp_channel *channel;
+	unsigned long timeout;
+	int err;
+
+	if (WARN_ON(irqs_disabled()))
+		return -EPERM;
+
+	if (!tegra_bpmp_message_valid(msg))
+		return -EINVAL;
+
+	channel = tegra_bpmp_write_threaded(bpmp, msg->mrq, msg->tx.data,
+					    msg->tx.size);
+	if (IS_ERR(channel))
+		return PTR_ERR(channel);
+
+	err = mbox_send_message(bpmp->mbox.channel, NULL);
+	if (err < 0)
+		return err;
+
+	mbox_client_txdone(bpmp->mbox.channel, 0);
+
+	timeout = usecs_to_jiffies(bpmp->soc->channels.thread.timeout);
+
+	err = wait_for_completion_timeout(&channel->completion, timeout);
+	if (err == 0)
+		return -ETIMEDOUT;
+
+	return tegra_bpmp_channel_read(channel, msg->rx.data, msg->rx.size);
+}
+EXPORT_SYMBOL_GPL(tegra_bpmp_transfer);
+
+static struct tegra_bpmp_mrq *tegra_bpmp_find_mrq(struct tegra_bpmp *bpmp,
+						  unsigned int mrq)
+{
+	struct tegra_bpmp_mrq *entry;
+
+	list_for_each_entry(entry, &bpmp->mrqs, list)
+		if (entry->mrq == mrq)
+			return entry;
+
+	return NULL;
+}
+
+static void tegra_bpmp_mrq_return(struct tegra_bpmp_channel *channel,
+				  int code, const void *data, size_t size)
+{
+	unsigned long flags = channel->ib->flags;
+	struct tegra_bpmp *bpmp = channel->bpmp;
+	struct tegra_bpmp_mb_data *frame;
+	int err;
+
+	if (WARN_ON(size > MSG_DATA_MIN_SZ))
+		return;
+
+	err = tegra_ivc_read_advance(channel->ivc);
+	if (WARN_ON(err < 0))
+		return;
+
+	if ((flags & MSG_ACK) == 0)
+		return;
+
+	frame = tegra_ivc_write_get_next_frame(channel->ivc);
+	if (WARN_ON(IS_ERR(frame)))
+		return;
+
+	frame->code = code;
+
+	if (data && size > 0)
+		memcpy(frame->data, data, size);
+
+	err = tegra_ivc_write_advance(channel->ivc);
+	if (WARN_ON(err < 0))
+		return;
+
+	if (flags & MSG_RING) {
+		err = mbox_send_message(bpmp->mbox.channel, NULL);
+		if (WARN_ON(err < 0))
+			return;
+
+		mbox_client_txdone(bpmp->mbox.channel, 0);
+	}
+}
+
+static void tegra_bpmp_handle_mrq(struct tegra_bpmp *bpmp,
+				  unsigned int mrq,
+				  struct tegra_bpmp_channel *channel)
+{
+	struct tegra_bpmp_mrq *entry;
+	u32 zero = 0;
+
+	spin_lock(&bpmp->lock);
+
+	entry = tegra_bpmp_find_mrq(bpmp, mrq);
+	if (!entry) {
+		spin_unlock(&bpmp->lock);
+		tegra_bpmp_mrq_return(channel, -EINVAL, &zero, sizeof(zero));
+		return;
+	}
+
+	entry->handler(mrq, channel, entry->data);
+
+	spin_unlock(&bpmp->lock);
+}
+
+int tegra_bpmp_request_mrq(struct tegra_bpmp *bpmp, unsigned int mrq,
+			   tegra_bpmp_mrq_handler_t handler, void *data)
+{
+	struct tegra_bpmp_mrq *entry;
+	unsigned long flags;
+
+	if (!handler)
+		return -EINVAL;
+
+	entry = devm_kzalloc(bpmp->dev, sizeof(*entry), GFP_KERNEL);
+	if (!entry)
+		return -ENOMEM;
+
+	spin_lock_irqsave(&bpmp->lock, flags);
+
+	entry->mrq = mrq;
+	entry->handler = handler;
+	entry->data = data;
+	list_add(&entry->list, &bpmp->mrqs);
+
+	spin_unlock_irqrestore(&bpmp->lock, flags);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(tegra_bpmp_request_mrq);
+
+void tegra_bpmp_free_mrq(struct tegra_bpmp *bpmp, unsigned int mrq, void *data)
+{
+	struct tegra_bpmp_mrq *entry;
+	unsigned long flags;
+
+	spin_lock_irqsave(&bpmp->lock, flags);
+
+	entry = tegra_bpmp_find_mrq(bpmp, mrq);
+	if (!entry)
+		goto unlock;
+
+	list_del(&entry->list);
+	devm_kfree(bpmp->dev, entry);
+
+unlock:
+	spin_unlock_irqrestore(&bpmp->lock, flags);
+}
+EXPORT_SYMBOL_GPL(tegra_bpmp_free_mrq);
+
+static void tegra_bpmp_mrq_handle_ping(unsigned int mrq,
+				       struct tegra_bpmp_channel *channel,
+				       void *data)
+{
+	struct mrq_ping_request *request;
+	struct mrq_ping_response response;
+
+	request = (struct mrq_ping_request *)channel->ib->data;
+
+	memset(&response, 0, sizeof(response));
+	response.reply = request->challenge << 1;
+
+	tegra_bpmp_mrq_return(channel, 0, &response, sizeof(response));
+}
+
+static int tegra_bpmp_ping(struct tegra_bpmp *bpmp)
+{
+	struct mrq_ping_response response;
+	struct mrq_ping_request request;
+	struct tegra_bpmp_message msg;
+	unsigned long flags;
+	ktime_t start, end;
+	int err;
+
+	memset(&request, 0, sizeof(request));
+	request.challenge = 1;
+
+	memset(&response, 0, sizeof(response));
+
+	memset(&msg, 0, sizeof(msg));
+	msg.mrq = MRQ_PING;
+	msg.tx.data = &request;
+	msg.tx.size = sizeof(request);
+	msg.rx.data = &response;
+	msg.rx.size = sizeof(response);
+
+	local_irq_save(flags);
+	start = ktime_get();
+	err = tegra_bpmp_transfer_atomic(bpmp, &msg);
+	end = ktime_get();
+	local_irq_restore(flags);
+
+	if (!err)
+		dev_dbg(bpmp->dev,
+			"ping ok: challenge: %u, response: %u, time: %lld\n",
+			request.challenge, response.reply,
+			ktime_to_us(ktime_sub(end, start)));
+
+	return err;
+}
+
+static int tegra_bpmp_get_firmware_tag(struct tegra_bpmp *bpmp, char *tag,
+				       size_t size)
+{
+	struct mrq_query_tag_request request;
+	struct tegra_bpmp_message msg;
+	unsigned long flags;
+	dma_addr_t phys;
+	void *virt;
+	int err;
+
+	virt = dma_alloc_coherent(bpmp->dev, MSG_DATA_MIN_SZ, &phys,
+				  GFP_KERNEL | GFP_DMA32);
+	if (!virt)
+		return -ENOMEM;
+
+	memset(&request, 0, sizeof(request));
+	request.addr = phys;
+
+	memset(&msg, 0, sizeof(msg));
+	msg.mrq = MRQ_QUERY_TAG;
+	msg.tx.data = &request;
+	msg.tx.size = sizeof(request);
+
+	local_irq_save(flags);
+	err = tegra_bpmp_transfer_atomic(bpmp, &msg);
+	local_irq_restore(flags);
+
+	if (err == 0)
+		strlcpy(tag, virt, size);
+
+	dma_free_coherent(bpmp->dev, MSG_DATA_MIN_SZ, virt, phys);
+
+	return err;
+}
+
+static void tegra_bpmp_channel_signal(struct tegra_bpmp_channel *channel)
+{
+	unsigned long flags = channel->ob->flags;
+
+	if ((flags & MSG_RING) == 0)
+		return;
+
+	complete(&channel->completion);
+}
+
+static void tegra_bpmp_handle_rx(struct mbox_client *client, void *data)
+{
+	struct tegra_bpmp *bpmp = mbox_client_to_bpmp(client);
+	struct tegra_bpmp_channel *channel;
+	unsigned int i, count;
+	unsigned long *busy;
+
+	channel = tegra_bpmp_channel_get_rx(bpmp);
+	count = bpmp->soc->channels.thread.count;
+	busy = bpmp->threaded.busy;
+
+	if (tegra_bpmp_master_acked(channel))
+		tegra_bpmp_handle_mrq(bpmp, channel->ib->code, channel);
+
+	spin_lock(&bpmp->lock);
+
+	for_each_set_bit(i, busy, count) {
+		struct tegra_bpmp_channel *channel;
+
+		channel = tegra_bpmp_channel_get_thread(bpmp, i);
+		if (!channel)
+			continue;
+
+		if (tegra_bpmp_master_acked(channel)) {
+			tegra_bpmp_channel_signal(channel);
+			clear_bit(i, busy);
+		}
+	}
+
+	spin_unlock(&bpmp->lock);
+}
+
+static void tegra_bpmp_ivc_notify(struct tegra_ivc *ivc, void *data)
+{
+	struct tegra_bpmp *bpmp = data;
+	int err;
+
+	if (WARN_ON(bpmp->mbox.channel == NULL))
+		return;
+
+	err = mbox_send_message(bpmp->mbox.channel, NULL);
+	if (err < 0)
+		return;
+
+	mbox_client_txdone(bpmp->mbox.channel, 0);
+}
+
+static int tegra_bpmp_channel_init(struct tegra_bpmp_channel *channel,
+				   struct tegra_bpmp *bpmp,
+				   unsigned int index)
+{
+	size_t message_size, queue_size;
+	unsigned int offset;
+	int err;
+
+	channel->ivc = devm_kzalloc(bpmp->dev, sizeof(*channel->ivc),
+				    GFP_KERNEL);
+	if (!channel->ivc)
+		return -ENOMEM;
+
+	message_size = tegra_ivc_align(MSG_MIN_SZ);
+	queue_size = tegra_ivc_total_queue_size(message_size);
+	offset = queue_size * index;
+
+	err = tegra_ivc_init(channel->ivc, NULL,
+			     bpmp->rx.virt + offset, bpmp->rx.phys + offset,
+			     bpmp->tx.virt + offset, bpmp->tx.phys + offset,
+			     1, message_size, tegra_bpmp_ivc_notify,
+			     bpmp);
+	if (err < 0) {
+		dev_err(bpmp->dev, "failed to setup IVC for channel %u: %d\n",
+			index, err);
+		return err;
+	}
+
+	init_completion(&channel->completion);
+	channel->bpmp = bpmp;
+
+	return 0;
+}
+
+static void tegra_bpmp_channel_reset(struct tegra_bpmp_channel *channel)
+{
+	/* reset the channel state */
+	tegra_ivc_reset(channel->ivc);
+
+	/* sync the channel state with BPMP */
+	while (tegra_ivc_notified(channel->ivc))
+		;
+}
+
+static void tegra_bpmp_channel_cleanup(struct tegra_bpmp_channel *channel)
+{
+	tegra_ivc_cleanup(channel->ivc);
+}
+
+static int tegra_bpmp_probe(struct platform_device *pdev)
+{
+	struct tegra_bpmp_channel *channel;
+	struct tegra_bpmp *bpmp;
+	unsigned int i;
+	char tag[32];
+	size_t size;
+	int err;
+
+	bpmp = devm_kzalloc(&pdev->dev, sizeof(*bpmp), GFP_KERNEL);
+	if (!bpmp)
+		return -ENOMEM;
+
+	bpmp->soc = of_device_get_match_data(&pdev->dev);
+	bpmp->dev = &pdev->dev;
+
+	bpmp->tx.pool = of_gen_pool_get(pdev->dev.of_node, "shmem", 0);
+	if (!bpmp->tx.pool) {
+		dev_err(&pdev->dev, "TX shmem pool not found\n");
+		return -ENOMEM;
+	}
+
+	bpmp->tx.virt = gen_pool_dma_alloc(bpmp->tx.pool, 4096, &bpmp->tx.phys);
+	if (!bpmp->tx.virt) {
+		dev_err(&pdev->dev, "failed to allocate from TX pool\n");
+		return -ENOMEM;
+	}
+
+	bpmp->rx.pool = of_gen_pool_get(pdev->dev.of_node, "shmem", 1);
+	if (!bpmp->rx.pool) {
+		dev_err(&pdev->dev, "RX shmem pool not found\n");
+		err = -ENOMEM;
+		goto free_tx;
+	}
+
+	bpmp->rx.virt = gen_pool_dma_alloc(bpmp->rx.pool, 4096, &bpmp->rx.phys);
+	if (!bpmp->rx.pool) {
+		dev_err(&pdev->dev, "failed to allocate from RX pool\n");
+		err = -ENOMEM;
+		goto free_tx;
+	}
+
+	INIT_LIST_HEAD(&bpmp->mrqs);
+	spin_lock_init(&bpmp->lock);
+
+	bpmp->threaded.count = bpmp->soc->channels.thread.count;
+	sema_init(&bpmp->threaded.lock, bpmp->threaded.count);
+
+	size = BITS_TO_LONGS(bpmp->threaded.count) * sizeof(long);
+
+	bpmp->threaded.allocated = devm_kzalloc(&pdev->dev, size, GFP_KERNEL);
+	if (!bpmp->threaded.allocated) {
+		err = -ENOMEM;
+		goto free_rx;
+	}
+
+	bpmp->threaded.busy = devm_kzalloc(&pdev->dev, size, GFP_KERNEL);
+	if (!bpmp->threaded.busy) {
+		err = -ENOMEM;
+		goto free_rx;
+	}
+
+	bpmp->num_channels = bpmp->soc->channels.cpu_tx.count +
+			     bpmp->soc->channels.thread.count +
+			     bpmp->soc->channels.cpu_rx.count;
+
+	bpmp->channels = devm_kcalloc(&pdev->dev, bpmp->num_channels,
+				      sizeof(*channel), GFP_KERNEL);
+	if (!bpmp->channels) {
+		err = -ENOMEM;
+		goto free_rx;
+	}
+
+	/* message channel initialization */
+	for (i = 0; i < bpmp->num_channels; i++) {
+		struct tegra_bpmp_channel *channel = &bpmp->channels[i];
+
+		err = tegra_bpmp_channel_init(channel, bpmp, i);
+		if (err < 0)
+			goto cleanup_channels;
+	}
+
+	/* mbox registration */
+	bpmp->mbox.client.dev = &pdev->dev;
+	bpmp->mbox.client.rx_callback = tegra_bpmp_handle_rx;
+	bpmp->mbox.client.tx_block = false;
+	bpmp->mbox.client.knows_txdone = false;
+
+	bpmp->mbox.channel = mbox_request_channel(&bpmp->mbox.client, 0);
+	if (IS_ERR(bpmp->mbox.channel)) {
+		err = PTR_ERR(bpmp->mbox.channel);
+		dev_err(&pdev->dev, "failed to get HSP mailbox: %d\n", err);
+		goto cleanup_channels;
+	}
+
+	/* reset message channels */
+	for (i = 0; i < bpmp->num_channels; i++) {
+		struct tegra_bpmp_channel *channel = &bpmp->channels[i];
+
+		tegra_bpmp_channel_reset(channel);
+	}
+
+	err = tegra_bpmp_request_mrq(bpmp, MRQ_PING,
+				     tegra_bpmp_mrq_handle_ping, bpmp);
+	if (err < 0)
+		goto free_mbox;
+
+	err = tegra_bpmp_ping(bpmp);
+	if (err < 0) {
+		dev_err(&pdev->dev, "failed to ping BPMP: %d\n", err);
+		goto free_mrq;
+	}
+
+	err = tegra_bpmp_get_firmware_tag(bpmp, tag, sizeof(tag) - 1);
+	if (err < 0) {
+		dev_err(&pdev->dev, "failed to get firmware tag: %d\n", err);
+		goto free_mrq;
+	}
+
+	dev_info(&pdev->dev, "firmware: %s\n", tag);
+
+	err = of_platform_default_populate(pdev->dev.of_node, NULL, &pdev->dev);
+	if (err < 0)
+		goto free_mrq;
+
+	err = tegra_bpmp_init_clocks(bpmp);
+	if (err < 0)
+		goto free_mrq;
+
+	err = tegra_bpmp_init_resets(bpmp);
+	if (err < 0)
+		goto free_mrq;
+
+	platform_set_drvdata(pdev, bpmp);
+
+	return 0;
+
+free_mrq:
+	tegra_bpmp_free_mrq(bpmp, MRQ_PING, bpmp);
+free_mbox:
+	mbox_free_channel(bpmp->mbox.channel);
+cleanup_channels:
+	while (i--)
+		tegra_bpmp_channel_cleanup(&bpmp->channels[i]);
+free_rx:
+	gen_pool_free(bpmp->rx.pool, (unsigned long)bpmp->rx.virt, 4096);
+free_tx:
+	gen_pool_free(bpmp->tx.pool, (unsigned long)bpmp->tx.virt, 4096);
+	return err;
+}
+
+static const struct tegra_bpmp_soc tegra186_soc = {
+	.channels = {
+		.cpu_tx = {
+			.offset = 0,
+			.count = 6,
+			.timeout = 60 * USEC_PER_SEC,
+		},
+		.thread = {
+			.offset = 6,
+			.count = 7,
+			.timeout = 600 * USEC_PER_SEC,
+		},
+		.cpu_rx = {
+			.offset = 13,
+			.count = 1,
+			.timeout = 0,
+		},
+	},
+	.num_resets = 193,
+};
+
+static const struct of_device_id tegra_bpmp_match[] = {
+	{ .compatible = "nvidia,tegra186-bpmp", .data = &tegra186_soc },
+	{ }
+};
+
+static struct platform_driver tegra_bpmp_driver = {
+	.driver = {
+		.name = "tegra-bpmp",
+		.of_match_table = tegra_bpmp_match,
+	},
+	.probe = tegra_bpmp_probe,
+};
+
+static int __init tegra_bpmp_init(void)
+{
+	return platform_driver_register(&tegra_bpmp_driver);
+}
+core_initcall(tegra_bpmp_init);
diff --git a/drivers/firmware/tegra/ivc.c b/drivers/firmware/tegra/ivc.c
new file mode 100644
index 0000000..29ecfd8
--- /dev/null
+++ b/drivers/firmware/tegra/ivc.c
@@ -0,0 +1,695 @@
+/*
+ * Copyright (c) 2014-2016, NVIDIA CORPORATION.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope 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 for
+ * more details.
+ */
+
+#include <soc/tegra/ivc.h>
+
+#define TEGRA_IVC_ALIGN 64
+
+/*
+ * IVC channel reset protocol.
+ *
+ * Each end uses its tx_channel.state to indicate its synchronization state.
+ */
+enum tegra_ivc_state {
+	/*
+	 * This value is zero for backwards compatibility with services that
+	 * assume channels to be initially zeroed. Such channels are in an
+	 * initially valid state, but cannot be asynchronously reset, and must
+	 * maintain a valid state at all times.
+	 *
+	 * The transmitting end can enter the established state from the sync or
+	 * ack state when it observes the receiving endpoint in the ack or
+	 * established state, indicating that has cleared the counters in our
+	 * rx_channel.
+	 */
+	TEGRA_IVC_STATE_ESTABLISHED = 0,
+
+	/*
+	 * If an endpoint is observed in the sync state, the remote endpoint is
+	 * allowed to clear the counters it owns asynchronously with respect to
+	 * the current endpoint. Therefore, the current endpoint is no longer
+	 * allowed to communicate.
+	 */
+	TEGRA_IVC_STATE_SYNC,
+
+	/*
+	 * When the transmitting end observes the receiving end in the sync
+	 * state, it can clear the w_count and r_count and transition to the ack
+	 * state. If the remote endpoint observes us in the ack state, it can
+	 * return to the established state once it has cleared its counters.
+	 */
+	TEGRA_IVC_STATE_ACK
+};
+
+/*
+ * This structure is divided into two-cache aligned parts, the first is only
+ * written through the tx.channel pointer, while the second is only written
+ * through the rx.channel pointer. This delineates ownership of the cache
+ * lines, which is critical to performance and necessary in non-cache coherent
+ * implementations.
+ */
+struct tegra_ivc_header {
+	union {
+		struct {
+			/* fields owned by the transmitting end */
+			u32 count;
+			u32 state;
+		};
+
+		u8 pad[TEGRA_IVC_ALIGN];
+	} tx;
+
+	union {
+		/* fields owned by the receiving end */
+		u32 count;
+		u8 pad[TEGRA_IVC_ALIGN];
+	} rx;
+};
+
+static inline void tegra_ivc_invalidate(struct tegra_ivc *ivc, dma_addr_t phys)
+{
+	if (!ivc->peer)
+		return;
+
+	dma_sync_single_for_cpu(ivc->peer, phys, TEGRA_IVC_ALIGN,
+				DMA_FROM_DEVICE);
+}
+
+static inline void tegra_ivc_flush(struct tegra_ivc *ivc, dma_addr_t phys)
+{
+	if (!ivc->peer)
+		return;
+
+	dma_sync_single_for_device(ivc->peer, phys, TEGRA_IVC_ALIGN,
+				   DMA_TO_DEVICE);
+}
+
+static inline bool tegra_ivc_empty(struct tegra_ivc *ivc,
+				   struct tegra_ivc_header *header)
+{
+	/*
+	 * This function performs multiple checks on the same values with
+	 * security implications, so create snapshots with ACCESS_ONCE() to
+	 * ensure that these checks use the same values.
+	 */
+	u32 tx = ACCESS_ONCE(header->tx.count);
+	u32 rx = ACCESS_ONCE(header->rx.count);
+
+	/*
+	 * Perform an over-full check to prevent denial of service attacks
+	 * where a server could be easily fooled into believing that there's
+	 * an extremely large number of frames ready, since receivers are not
+	 * expected to check for full or over-full conditions.
+	 *
+	 * Although the channel isn't empty, this is an invalid case caused by
+	 * a potentially malicious peer, so returning empty is safer, because
+	 * it gives the impression that the channel has gone silent.
+	 */
+	if (tx - rx > ivc->num_frames)
+		return true;
+
+	return tx == rx;
+}
+
+static inline bool tegra_ivc_full(struct tegra_ivc *ivc,
+				  struct tegra_ivc_header *header)
+{
+	u32 tx = ACCESS_ONCE(header->tx.count);
+	u32 rx = ACCESS_ONCE(header->rx.count);
+
+	/*
+	 * Invalid cases where the counters indicate that the queue is over
+	 * capacity also appear full.
+	 */
+	return tx - rx >= ivc->num_frames;
+}
+
+static inline u32 tegra_ivc_available(struct tegra_ivc *ivc,
+				      struct tegra_ivc_header *header)
+{
+	u32 tx = ACCESS_ONCE(header->tx.count);
+	u32 rx = ACCESS_ONCE(header->rx.count);
+
+	/*
+	 * This function isn't expected to be used in scenarios where an
+	 * over-full situation can lead to denial of service attacks. See the
+	 * comment in tegra_ivc_empty() for an explanation about special
+	 * over-full considerations.
+	 */
+	return tx - rx;
+}
+
+static inline void tegra_ivc_advance_tx(struct tegra_ivc *ivc)
+{
+	ACCESS_ONCE(ivc->tx.channel->tx.count) =
+		ACCESS_ONCE(ivc->tx.channel->tx.count) + 1;
+
+	if (ivc->tx.position == ivc->num_frames - 1)
+		ivc->tx.position = 0;
+	else
+		ivc->tx.position++;
+}
+
+static inline void tegra_ivc_advance_rx(struct tegra_ivc *ivc)
+{
+	ACCESS_ONCE(ivc->rx.channel->rx.count) =
+		ACCESS_ONCE(ivc->rx.channel->rx.count) + 1;
+
+	if (ivc->rx.position == ivc->num_frames - 1)
+		ivc->rx.position = 0;
+	else
+		ivc->rx.position++;
+}
+
+static inline int tegra_ivc_check_read(struct tegra_ivc *ivc)
+{
+	unsigned int offset = offsetof(struct tegra_ivc_header, tx.count);
+
+	/*
+	 * tx.channel->state is set locally, so it is not synchronized with
+	 * state from the remote peer. The remote peer cannot reset its
+	 * transmit counters until we've acknowledged its synchronization
+	 * request, so no additional synchronization is required because an
+	 * asynchronous transition of rx.channel->state to
+	 * TEGRA_IVC_STATE_ACK is not allowed.
+	 */
+	if (ivc->tx.channel->tx.state != TEGRA_IVC_STATE_ESTABLISHED)
+		return -ECONNRESET;
+
+	/*
+	 * Avoid unnecessary invalidations when performing repeated accesses
+	 * to an IVC channel by checking the old queue pointers first.
+	 *
+	 * Synchronization is only necessary when these pointers indicate
+	 * empty or full.
+	 */
+	if (!tegra_ivc_empty(ivc, ivc->rx.channel))
+		return 0;
+
+	tegra_ivc_invalidate(ivc, ivc->rx.phys + offset);
+
+	if (tegra_ivc_empty(ivc, ivc->rx.channel))
+		return -ENOSPC;
+
+	return 0;
+}
+
+static inline int tegra_ivc_check_write(struct tegra_ivc *ivc)
+{
+	unsigned int offset = offsetof(struct tegra_ivc_header, rx.count);
+
+	if (ivc->tx.channel->tx.state != TEGRA_IVC_STATE_ESTABLISHED)
+		return -ECONNRESET;
+
+	if (!tegra_ivc_full(ivc, ivc->tx.channel))
+		return 0;
+
+	tegra_ivc_invalidate(ivc, ivc->tx.phys + offset);
+
+	if (tegra_ivc_full(ivc, ivc->tx.channel))
+		return -ENOSPC;
+
+	return 0;
+}
+
+static void *tegra_ivc_frame_virt(struct tegra_ivc *ivc,
+				  struct tegra_ivc_header *header,
+				  unsigned int frame)
+{
+	if (WARN_ON(frame >= ivc->num_frames))
+		return ERR_PTR(-EINVAL);
+
+	return (void *)(header + 1) + ivc->frame_size * frame;
+}
+
+static inline dma_addr_t tegra_ivc_frame_phys(struct tegra_ivc *ivc,
+					      dma_addr_t phys,
+					      unsigned int frame)
+{
+	unsigned long offset;
+
+	offset = sizeof(struct tegra_ivc_header) + ivc->frame_size * frame;
+
+	return phys + offset;
+}
+
+static inline void tegra_ivc_invalidate_frame(struct tegra_ivc *ivc,
+					      dma_addr_t phys,
+					      unsigned int frame,
+					      unsigned int offset,
+					      size_t size)
+{
+	if (!ivc->peer || WARN_ON(frame >= ivc->num_frames))
+		return;
+
+	phys = tegra_ivc_frame_phys(ivc, phys, frame) + offset;
+
+	dma_sync_single_for_cpu(ivc->peer, phys, size, DMA_FROM_DEVICE);
+}
+
+static inline void tegra_ivc_flush_frame(struct tegra_ivc *ivc,
+					 dma_addr_t phys,
+					 unsigned int frame,
+					 unsigned int offset,
+					 size_t size)
+{
+	if (!ivc->peer || WARN_ON(frame >= ivc->num_frames))
+		return;
+
+	phys = tegra_ivc_frame_phys(ivc, phys, frame) + offset;
+
+	dma_sync_single_for_device(ivc->peer, phys, size, DMA_TO_DEVICE);
+}
+
+/* directly peek at the next frame rx'ed */
+void *tegra_ivc_read_get_next_frame(struct tegra_ivc *ivc)
+{
+	int err;
+
+	if (WARN_ON(ivc == NULL))
+		return ERR_PTR(-EINVAL);
+
+	err = tegra_ivc_check_read(ivc);
+	if (err < 0)
+		return ERR_PTR(err);
+
+	/*
+	 * Order observation of ivc->rx.position potentially indicating new
+	 * data before data read.
+	 */
+	smp_rmb();
+
+	tegra_ivc_invalidate_frame(ivc, ivc->rx.phys, ivc->rx.position, 0,
+				   ivc->frame_size);
+
+	return tegra_ivc_frame_virt(ivc, ivc->rx.channel, ivc->rx.position);
+}
+EXPORT_SYMBOL(tegra_ivc_read_get_next_frame);
+
+int tegra_ivc_read_advance(struct tegra_ivc *ivc)
+{
+	unsigned int rx = offsetof(struct tegra_ivc_header, rx.count);
+	unsigned int tx = offsetof(struct tegra_ivc_header, tx.count);
+	int err;
+
+	/*
+	 * No read barriers or synchronization here: the caller is expected to
+	 * have already observed the channel non-empty. This check is just to
+	 * catch programming errors.
+	 */
+	err = tegra_ivc_check_read(ivc);
+	if (err < 0)
+		return err;
+
+	tegra_ivc_advance_rx(ivc);
+
+	tegra_ivc_flush(ivc, ivc->rx.phys + rx);
+
+	/*
+	 * Ensure our write to ivc->rx.position occurs before our read from
+	 * ivc->tx.position.
+	 */
+	smp_mb();
+
+	/*
+	 * Notify only upon transition from full to non-full. The available
+	 * count can only asynchronously increase, so the worst possible
+	 * side-effect will be a spurious notification.
+	 */
+	tegra_ivc_invalidate(ivc, ivc->rx.phys + tx);
+
+	if (tegra_ivc_available(ivc, ivc->rx.channel) == ivc->num_frames - 1)
+		ivc->notify(ivc, ivc->notify_data);
+
+	return 0;
+}
+EXPORT_SYMBOL(tegra_ivc_read_advance);
+
+/* directly poke at the next frame to be tx'ed */
+void *tegra_ivc_write_get_next_frame(struct tegra_ivc *ivc)
+{
+	int err;
+
+	err = tegra_ivc_check_write(ivc);
+	if (err < 0)
+		return ERR_PTR(err);
+
+	return tegra_ivc_frame_virt(ivc, ivc->tx.channel, ivc->tx.position);
+}
+EXPORT_SYMBOL(tegra_ivc_write_get_next_frame);
+
+/* advance the tx buffer */
+int tegra_ivc_write_advance(struct tegra_ivc *ivc)
+{
+	unsigned int tx = offsetof(struct tegra_ivc_header, tx.count);
+	unsigned int rx = offsetof(struct tegra_ivc_header, rx.count);
+	int err;
+
+	err = tegra_ivc_check_write(ivc);
+	if (err < 0)
+		return err;
+
+	tegra_ivc_flush_frame(ivc, ivc->tx.phys, ivc->tx.position, 0,
+			      ivc->frame_size);
+
+	/*
+	 * Order any possible stores to the frame before update of
+	 * ivc->tx.position.
+	 */
+	smp_wmb();
+
+	tegra_ivc_advance_tx(ivc);
+	tegra_ivc_flush(ivc, ivc->tx.phys + tx);
+
+	/*
+	 * Ensure our write to ivc->tx.position occurs before our read from
+	 * ivc->rx.position.
+	 */
+	smp_mb();
+
+	/*
+	 * Notify only upon transition from empty to non-empty. The available
+	 * count can only asynchronously decrease, so the worst possible
+	 * side-effect will be a spurious notification.
+	 */
+	tegra_ivc_invalidate(ivc, ivc->tx.phys + rx);
+
+	if (tegra_ivc_available(ivc, ivc->tx.channel) == 1)
+		ivc->notify(ivc, ivc->notify_data);
+
+	return 0;
+}
+EXPORT_SYMBOL(tegra_ivc_write_advance);
+
+void tegra_ivc_reset(struct tegra_ivc *ivc)
+{
+	unsigned int offset = offsetof(struct tegra_ivc_header, tx.count);
+
+	ivc->tx.channel->tx.state = TEGRA_IVC_STATE_SYNC;
+	tegra_ivc_flush(ivc, ivc->tx.phys + offset);
+	ivc->notify(ivc, ivc->notify_data);
+}
+EXPORT_SYMBOL(tegra_ivc_reset);
+
+/*
+ * =======================================================
+ *  IVC State Transition Table - see tegra_ivc_notified()
+ * =======================================================
+ *
+ *	local	remote	action
+ *	-----	------	-----------------------------------
+ *	SYNC	EST	<none>
+ *	SYNC	ACK	reset counters; move to EST; notify
+ *	SYNC	SYNC	reset counters; move to ACK; notify
+ *	ACK	EST	move to EST; notify
+ *	ACK	ACK	move to EST; notify
+ *	ACK	SYNC	reset counters; move to ACK; notify
+ *	EST	EST	<none>
+ *	EST	ACK	<none>
+ *	EST	SYNC	reset counters; move to ACK; notify
+ *
+ * ===============================================================
+ */
+
+int tegra_ivc_notified(struct tegra_ivc *ivc)
+{
+	unsigned int offset = offsetof(struct tegra_ivc_header, tx.count);
+	enum tegra_ivc_state state;
+
+	/* Copy the receiver's state out of shared memory. */
+	tegra_ivc_invalidate(ivc, ivc->rx.phys + offset);
+	state = ACCESS_ONCE(ivc->rx.channel->tx.state);
+
+	if (state == TEGRA_IVC_STATE_SYNC) {
+		offset = offsetof(struct tegra_ivc_header, tx.count);
+
+		/*
+		 * Order observation of TEGRA_IVC_STATE_SYNC before stores
+		 * clearing tx.channel.
+		 */
+		smp_rmb();
+
+		/*
+		 * Reset tx.channel counters. The remote end is in the SYNC
+		 * state and won't make progress until we change our state,
+		 * so the counters are not in use at this time.
+		 */
+		ivc->tx.channel->tx.count = 0;
+		ivc->rx.channel->rx.count = 0;
+
+		ivc->tx.position = 0;
+		ivc->rx.position = 0;
+
+		/*
+		 * Ensure that counters appear cleared before new state can be
+		 * observed.
+		 */
+		smp_wmb();
+
+		/*
+		 * Move to ACK state. We have just cleared our counters, so it
+		 * is now safe for the remote end to start using these values.
+		 */
+		ivc->tx.channel->tx.state = TEGRA_IVC_STATE_ACK;
+		tegra_ivc_flush(ivc, ivc->tx.phys + offset);
+
+		/*
+		 * Notify remote end to observe state transition.
+		 */
+		ivc->notify(ivc, ivc->notify_data);
+
+	} else if (ivc->tx.channel->tx.state == TEGRA_IVC_STATE_SYNC &&
+		   state == TEGRA_IVC_STATE_ACK) {
+		offset = offsetof(struct tegra_ivc_header, tx.count);
+
+		/*
+		 * Order observation of ivc_state_sync before stores clearing
+		 * tx_channel.
+		 */
+		smp_rmb();
+
+		/*
+		 * Reset tx.channel counters. The remote end is in the ACK
+		 * state and won't make progress until we change our state,
+		 * so the counters are not in use at this time.
+		 */
+		ivc->tx.channel->tx.count = 0;
+		ivc->rx.channel->rx.count = 0;
+
+		ivc->tx.position = 0;
+		ivc->rx.position = 0;
+
+		/*
+		 * Ensure that counters appear cleared before new state can be
+		 * observed.
+		 */
+		smp_wmb();
+
+		/*
+		 * Move to ESTABLISHED state. We know that the remote end has
+		 * already cleared its counters, so it is safe to start
+		 * writing/reading on this channel.
+		 */
+		ivc->tx.channel->tx.state = TEGRA_IVC_STATE_ESTABLISHED;
+		tegra_ivc_flush(ivc, ivc->tx.phys + offset);
+
+		/*
+		 * Notify remote end to observe state transition.
+		 */
+		ivc->notify(ivc, ivc->notify_data);
+
+	} else if (ivc->tx.channel->tx.state == TEGRA_IVC_STATE_ACK) {
+		offset = offsetof(struct tegra_ivc_header, tx.count);
+
+		/*
+		 * At this point, we have observed the peer to be in either
+		 * the ACK or ESTABLISHED state. Next, order observation of
+		 * peer state before storing to tx.channel.
+		 */
+		smp_rmb();
+
+		/*
+		 * Move to ESTABLISHED state. We know that we have previously
+		 * cleared our counters, and we know that the remote end has
+		 * cleared its counters, so it is safe to start writing/reading
+		 * on this channel.
+		 */
+		ivc->tx.channel->tx.state = TEGRA_IVC_STATE_ESTABLISHED;
+		tegra_ivc_flush(ivc, ivc->tx.phys + offset);
+
+		/*
+		 * Notify remote end to observe state transition.
+		 */
+		ivc->notify(ivc, ivc->notify_data);
+
+	} else {
+		/*
+		 * There is no need to handle any further action. Either the
+		 * channel is already fully established, or we are waiting for
+		 * the remote end to catch up with our current state. Refer
+		 * to the diagram in "IVC State Transition Table" above.
+		 */
+	}
+
+	if (ivc->tx.channel->tx.state != TEGRA_IVC_STATE_ESTABLISHED)
+		return -EAGAIN;
+
+	return 0;
+}
+EXPORT_SYMBOL(tegra_ivc_notified);
+
+size_t tegra_ivc_align(size_t size)
+{
+	return ALIGN(size, TEGRA_IVC_ALIGN);
+}
+EXPORT_SYMBOL(tegra_ivc_align);
+
+unsigned tegra_ivc_total_queue_size(unsigned queue_size)
+{
+	if (!IS_ALIGNED(queue_size, TEGRA_IVC_ALIGN)) {
+		pr_err("%s: queue_size (%u) must be %u-byte aligned\n",
+		       __func__, queue_size, TEGRA_IVC_ALIGN);
+		return 0;
+	}
+
+	return queue_size + sizeof(struct tegra_ivc_header);
+}
+EXPORT_SYMBOL(tegra_ivc_total_queue_size);
+
+static int tegra_ivc_check_params(unsigned long rx, unsigned long tx,
+				  unsigned int num_frames, size_t frame_size)
+{
+	BUILD_BUG_ON(!IS_ALIGNED(offsetof(struct tegra_ivc_header, tx.count),
+				 TEGRA_IVC_ALIGN));
+	BUILD_BUG_ON(!IS_ALIGNED(offsetof(struct tegra_ivc_header, rx.count),
+				 TEGRA_IVC_ALIGN));
+	BUILD_BUG_ON(!IS_ALIGNED(sizeof(struct tegra_ivc_header),
+				 TEGRA_IVC_ALIGN));
+
+	if ((uint64_t)num_frames * (uint64_t)frame_size >= 0x100000000UL) {
+		pr_err("num_frames * frame_size overflows\n");
+		return -EINVAL;
+	}
+
+	if (!IS_ALIGNED(frame_size, TEGRA_IVC_ALIGN)) {
+		pr_err("frame size not adequately aligned: %zu\n", frame_size);
+		return -EINVAL;
+	}
+
+	/*
+	 * The headers must at least be aligned enough for counters
+	 * to be accessed atomically.
+	 */
+	if (!IS_ALIGNED(rx, TEGRA_IVC_ALIGN)) {
+		pr_err("IVC channel start not aligned: %#lx\n", rx);
+		return -EINVAL;
+	}
+
+	if (!IS_ALIGNED(tx, TEGRA_IVC_ALIGN)) {
+		pr_err("IVC channel start not aligned: %#lx\n", tx);
+		return -EINVAL;
+	}
+
+	if (rx < tx) {
+		if (rx + frame_size * num_frames > tx) {
+			pr_err("queue regions overlap: %#lx + %zx > %#lx\n",
+			       rx, frame_size * num_frames, tx);
+			return -EINVAL;
+		}
+	} else {
+		if (tx + frame_size * num_frames > rx) {
+			pr_err("queue regions overlap: %#lx + %zx > %#lx\n",
+			       tx, frame_size * num_frames, rx);
+			return -EINVAL;
+		}
+	}
+
+	return 0;
+}
+
+int tegra_ivc_init(struct tegra_ivc *ivc, struct device *peer, void *rx,
+		   dma_addr_t rx_phys, void *tx, dma_addr_t tx_phys,
+		   unsigned int num_frames, size_t frame_size,
+		   void (*notify)(struct tegra_ivc *ivc, void *data),
+		   void *data)
+{
+	size_t queue_size;
+	int err;
+
+	if (WARN_ON(!ivc || !notify))
+		return -EINVAL;
+
+	/*
+	 * All sizes that can be returned by communication functions should
+	 * fit in an int.
+	 */
+	if (frame_size > INT_MAX)
+		return -E2BIG;
+
+	err = tegra_ivc_check_params((unsigned long)rx, (unsigned long)tx,
+				     num_frames, frame_size);
+	if (err < 0)
+		return err;
+
+	queue_size = tegra_ivc_total_queue_size(num_frames * frame_size);
+
+	if (peer) {
+		ivc->rx.phys = dma_map_single(peer, rx, queue_size,
+					      DMA_BIDIRECTIONAL);
+		if (ivc->rx.phys == DMA_ERROR_CODE)
+			return -ENOMEM;
+
+		ivc->tx.phys = dma_map_single(peer, tx, queue_size,
+					      DMA_BIDIRECTIONAL);
+		if (ivc->tx.phys == DMA_ERROR_CODE) {
+			dma_unmap_single(peer, ivc->rx.phys, queue_size,
+					 DMA_BIDIRECTIONAL);
+			return -ENOMEM;
+		}
+	} else {
+		ivc->rx.phys = rx_phys;
+		ivc->tx.phys = tx_phys;
+	}
+
+	ivc->rx.channel = rx;
+	ivc->tx.channel = tx;
+	ivc->peer = peer;
+	ivc->notify = notify;
+	ivc->notify_data = data;
+	ivc->frame_size = frame_size;
+	ivc->num_frames = num_frames;
+
+	/*
+	 * These values aren't necessarily correct until the channel has been
+	 * reset.
+	 */
+	ivc->tx.position = 0;
+	ivc->rx.position = 0;
+
+	return 0;
+}
+EXPORT_SYMBOL(tegra_ivc_init);
+
+void tegra_ivc_cleanup(struct tegra_ivc *ivc)
+{
+	if (ivc->peer) {
+		size_t size = tegra_ivc_total_queue_size(ivc->num_frames *
+							 ivc->frame_size);
+
+		dma_unmap_single(ivc->peer, ivc->rx.phys, size,
+				 DMA_BIDIRECTIONAL);
+		dma_unmap_single(ivc->peer, ivc->tx.phys, size,
+				 DMA_BIDIRECTIONAL);
+	}
+}
+EXPORT_SYMBOL(tegra_ivc_cleanup);