summary refs log tree commit diff
diff options
context:
space:
mode:
-rw-r--r--Documentation/ABI/testing/sysfs-bus-rpmsg75
-rw-r--r--Documentation/rpmsg.txt293
-rw-r--r--drivers/Kconfig2
-rw-r--r--drivers/Makefile1
-rw-r--r--drivers/rpmsg/Kconfig5
-rw-r--r--drivers/rpmsg/Makefile1
-rw-r--r--drivers/rpmsg/virtio_rpmsg_bus.c1026
-rw-r--r--include/linux/mod_devicetable.h9
-rw-r--r--include/linux/rpmsg.h326
-rw-r--r--include/linux/virtio_ids.h1
10 files changed, 1739 insertions, 0 deletions
diff --git a/Documentation/ABI/testing/sysfs-bus-rpmsg b/Documentation/ABI/testing/sysfs-bus-rpmsg
new file mode 100644
index 000000000000..189e419a5a2d
--- /dev/null
+++ b/Documentation/ABI/testing/sysfs-bus-rpmsg
@@ -0,0 +1,75 @@
+What:		/sys/bus/rpmsg/devices/.../name
+Date:		June 2011
+KernelVersion:	3.3
+Contact:	Ohad Ben-Cohen <ohad@wizery.com>
+Description:
+		Every rpmsg device is a communication channel with a remote
+		processor. Channels are identified with a (textual) name,
+		which is maximum 32 bytes long (defined as RPMSG_NAME_SIZE in
+		rpmsg.h).
+
+		This sysfs entry contains the name of this channel.
+
+What:		/sys/bus/rpmsg/devices/.../src
+Date:		June 2011
+KernelVersion:	3.3
+Contact:	Ohad Ben-Cohen <ohad@wizery.com>
+Description:
+		Every rpmsg device is a communication channel with a remote
+		processor. Channels have a local ("source") rpmsg address,
+		and remote ("destination") rpmsg address. When an entity
+		starts listening on one end of a channel, it assigns it with
+		a unique rpmsg address (a 32 bits integer). This way when
+		inbound messages arrive to this address, the rpmsg core
+		dispatches them to the listening entity (a kernel driver).
+
+		This sysfs entry contains the src (local) rpmsg address
+		of this channel. If it contains 0xffffffff, then an address
+		wasn't assigned (can happen if no driver exists for this
+		channel).
+
+What:		/sys/bus/rpmsg/devices/.../dst
+Date:		June 2011
+KernelVersion:	3.3
+Contact:	Ohad Ben-Cohen <ohad@wizery.com>
+Description:
+		Every rpmsg device is a communication channel with a remote
+		processor. Channels have a local ("source") rpmsg address,
+		and remote ("destination") rpmsg address. When an entity
+		starts listening on one end of a channel, it assigns it with
+		a unique rpmsg address (a 32 bits integer). This way when
+		inbound messages arrive to this address, the rpmsg core
+		dispatches them to the listening entity.
+
+		This sysfs entry contains the dst (remote) rpmsg address
+		of this channel. If it contains 0xffffffff, then an address
+		wasn't assigned (can happen if the kernel driver that
+		is attached to this channel is exposing a service to the
+		remote processor. This make it a local rpmsg server,
+		and it is listening for inbound messages that may be sent
+		from any remote rpmsg client; it is not bound to a single
+		remote entity).
+
+What:		/sys/bus/rpmsg/devices/.../announce
+Date:		June 2011
+KernelVersion:	3.3
+Contact:	Ohad Ben-Cohen <ohad@wizery.com>
+Description:
+		Every rpmsg device is a communication channel with a remote
+		processor. Channels are identified by a textual name (see
+		/sys/bus/rpmsg/devices/.../name above) and have a local
+		("source") rpmsg address, and remote ("destination") rpmsg
+		address.
+
+		A channel is first created when an entity, whether local
+		or remote, starts listening on it for messages (and is thus
+		called an rpmsg server).
+
+		When that happens, a "name service" announcement is sent
+		to the other processor, in order to let it know about the
+		creation of the channel (this way remote clients know they
+		can start sending messages).
+
+		This sysfs entry tells us whether the channel is a local
+		server channel that is announced (values are either
+		true or false).
diff --git a/Documentation/rpmsg.txt b/Documentation/rpmsg.txt
new file mode 100644
index 000000000000..409d9f964c5b
--- /dev/null
+++ b/Documentation/rpmsg.txt
@@ -0,0 +1,293 @@
+Remote Processor Messaging (rpmsg) Framework
+
+Note: this document describes the rpmsg bus and how to write rpmsg drivers.
+To learn how to add rpmsg support for new platforms, check out remoteproc.txt
+(also a resident of Documentation/).
+
+1. Introduction
+
+Modern SoCs typically employ heterogeneous remote processor devices in
+asymmetric multiprocessing (AMP) configurations, which may be running
+different instances of operating system, whether it's Linux or any other
+flavor of real-time OS.
+
+OMAP4, for example, has dual Cortex-A9, dual Cortex-M3 and a C64x+ DSP.
+Typically, the dual cortex-A9 is running Linux in a SMP configuration,
+and each of the other three cores (two M3 cores and a DSP) is running
+its own instance of RTOS in an AMP configuration.
+
+Typically AMP remote processors employ dedicated DSP codecs and multimedia
+hardware accelerators, and therefore are often used to offload CPU-intensive
+multimedia tasks from the main application processor.
+
+These remote processors could also be used to control latency-sensitive
+sensors, drive random hardware blocks, or just perform background tasks
+while the main CPU is idling.
+
+Users of those remote processors can either be userland apps (e.g. multimedia
+frameworks talking with remote OMX components) or kernel drivers (controlling
+hardware accessible only by the remote processor, reserving kernel-controlled
+resources on behalf of the remote processor, etc..).
+
+Rpmsg is a virtio-based messaging bus that allows kernel drivers to communicate
+with remote processors available on the system. In turn, drivers could then
+expose appropriate user space interfaces, if needed.
+
+When writing a driver that exposes rpmsg communication to userland, please
+keep in mind that remote processors might have direct access to the
+system's physical memory and other sensitive hardware resources (e.g. on
+OMAP4, remote cores and hardware accelerators may have direct access to the
+physical memory, gpio banks, dma controllers, i2c bus, gptimers, mailbox
+devices, hwspinlocks, etc..). Moreover, those remote processors might be
+running RTOS where every task can access the entire memory/devices exposed
+to the processor. To minimize the risks of rogue (or buggy) userland code
+exploiting remote bugs, and by that taking over the system, it is often
+desired to limit userland to specific rpmsg channels (see definition below)
+it can send messages on, and if possible, minimize how much control
+it has over the content of the messages.
+
+Every rpmsg device is a communication channel with a remote processor (thus
+rpmsg devices are called channels). Channels are identified by a textual name
+and have a local ("source") rpmsg address, and remote ("destination") rpmsg
+address.
+
+When a driver starts listening on a channel, its rx callback is bound with
+a unique rpmsg local address (a 32-bit integer). This way when inbound messages
+arrive, the rpmsg core dispatches them to the appropriate driver according
+to their destination address (this is done by invoking the driver's rx handler
+with the payload of the inbound message).
+
+
+2. User API
+
+  int rpmsg_send(struct rpmsg_channel *rpdev, void *data, int len);
+   - sends a message across to the remote processor on a given channel.
+     The caller should specify the channel, the data it wants to send,
+     and its length (in bytes). The message will be sent on the specified
+     channel, i.e. its source and destination address fields will be
+     set to the channel's src and dst addresses.
+
+     In case there are no TX buffers available, the function will block until
+     one becomes available (i.e. until the remote processor consumes
+     a tx buffer and puts it back on virtio's used descriptor ring),
+     or a timeout of 15 seconds elapses. When the latter happens,
+     -ERESTARTSYS is returned.
+     The function can only be called from a process context (for now).
+     Returns 0 on success and an appropriate error value on failure.
+
+  int rpmsg_sendto(struct rpmsg_channel *rpdev, void *data, int len, u32 dst);
+   - sends a message across to the remote processor on a given channel,
+     to a destination address provided by the caller.
+     The caller should specify the channel, the data it wants to send,
+     its length (in bytes), and an explicit destination address.
+     The message will then be sent to the remote processor to which the
+     channel belongs, using the channel's src address, and the user-provided
+     dst address (thus the channel's dst address will be ignored).
+
+     In case there are no TX buffers available, the function will block until
+     one becomes available (i.e. until the remote processor consumes
+     a tx buffer and puts it back on virtio's used descriptor ring),
+     or a timeout of 15 seconds elapses. When the latter happens,
+     -ERESTARTSYS is returned.
+     The function can only be called from a process context (for now).
+     Returns 0 on success and an appropriate error value on failure.
+
+  int rpmsg_send_offchannel(struct rpmsg_channel *rpdev, u32 src, u32 dst,
+							void *data, int len);
+   - sends a message across to the remote processor, using the src and dst
+     addresses provided by the user.
+     The caller should specify the channel, the data it wants to send,
+     its length (in bytes), and explicit source and destination addresses.
+     The message will then be sent to the remote processor to which the
+     channel belongs, but the channel's src and dst addresses will be
+     ignored (and the user-provided addresses will be used instead).
+
+     In case there are no TX buffers available, the function will block until
+     one becomes available (i.e. until the remote processor consumes
+     a tx buffer and puts it back on virtio's used descriptor ring),
+     or a timeout of 15 seconds elapses. When the latter happens,
+     -ERESTARTSYS is returned.
+     The function can only be called from a process context (for now).
+     Returns 0 on success and an appropriate error value on failure.
+
+  int rpmsg_trysend(struct rpmsg_channel *rpdev, void *data, int len);
+   - sends a message across to the remote processor on a given channel.
+     The caller should specify the channel, the data it wants to send,
+     and its length (in bytes). The message will be sent on the specified
+     channel, i.e. its source and destination address fields will be
+     set to the channel's src and dst addresses.
+
+     In case there are no TX buffers available, the function will immediately
+     return -ENOMEM without waiting until one becomes available.
+     The function can only be called from a process context (for now).
+     Returns 0 on success and an appropriate error value on failure.
+
+  int rpmsg_trysendto(struct rpmsg_channel *rpdev, void *data, int len, u32 dst)
+   - sends a message across to the remote processor on a given channel,
+     to a destination address provided by the user.
+     The user should specify the channel, the data it wants to send,
+     its length (in bytes), and an explicit destination address.
+     The message will then be sent to the remote processor to which the
+     channel belongs, using the channel's src address, and the user-provided
+     dst address (thus the channel's dst address will be ignored).
+
+     In case there are no TX buffers available, the function will immediately
+     return -ENOMEM without waiting until one becomes available.
+     The function can only be called from a process context (for now).
+     Returns 0 on success and an appropriate error value on failure.
+
+  int rpmsg_trysend_offchannel(struct rpmsg_channel *rpdev, u32 src, u32 dst,
+							void *data, int len);
+   - sends a message across to the remote processor, using source and
+     destination addresses provided by the user.
+     The user should specify the channel, the data it wants to send,
+     its length (in bytes), and explicit source and destination addresses.
+     The message will then be sent to the remote processor to which the
+     channel belongs, but the channel's src and dst addresses will be
+     ignored (and the user-provided addresses will be used instead).
+
+     In case there are no TX buffers available, the function will immediately
+     return -ENOMEM without waiting until one becomes available.
+     The function can only be called from a process context (for now).
+     Returns 0 on success and an appropriate error value on failure.
+
+  struct rpmsg_endpoint *rpmsg_create_ept(struct rpmsg_channel *rpdev,
+		void (*cb)(struct rpmsg_channel *, void *, int, void *, u32),
+		void *priv, u32 addr);
+   - every rpmsg address in the system is bound to an rx callback (so when
+     inbound messages arrive, they are dispatched by the rpmsg bus using the
+     appropriate callback handler) by means of an rpmsg_endpoint struct.
+
+     This function allows drivers to create such an endpoint, and by that,
+     bind a callback, and possibly some private data too, to an rpmsg address
+     (either one that is known in advance, or one that will be dynamically
+     assigned for them).
+
+     Simple rpmsg drivers need not call rpmsg_create_ept, because an endpoint
+     is already created for them when they are probed by the rpmsg bus
+     (using the rx callback they provide when they registered to the rpmsg bus).
+
+     So things should just work for simple drivers: they already have an
+     endpoint, their rx callback is bound to their rpmsg address, and when
+     relevant inbound messages arrive (i.e. messages which their dst address
+     equals to the src address of their rpmsg channel), the driver's handler
+     is invoked to process it.
+
+     That said, more complicated drivers might do need to allocate
+     additional rpmsg addresses, and bind them to different rx callbacks.
+     To accomplish that, those drivers need to call this function.
+     Drivers should provide their channel (so the new endpoint would bind
+     to the same remote processor their channel belongs to), an rx callback
+     function, an optional private data (which is provided back when the
+     rx callback is invoked), and an address they want to bind with the
+     callback. If addr is RPMSG_ADDR_ANY, then rpmsg_create_ept will
+     dynamically assign them an available rpmsg address (drivers should have
+     a very good reason why not to always use RPMSG_ADDR_ANY here).
+
+     Returns a pointer to the endpoint on success, or NULL on error.
+
+  void rpmsg_destroy_ept(struct rpmsg_endpoint *ept);
+   - destroys an existing rpmsg endpoint. user should provide a pointer
+     to an rpmsg endpoint that was previously created with rpmsg_create_ept().
+
+  int register_rpmsg_driver(struct rpmsg_driver *rpdrv);
+   - registers an rpmsg driver with the rpmsg bus. user should provide
+     a pointer to an rpmsg_driver struct, which contains the driver's
+     ->probe() and ->remove() functions, an rx callback, and an id_table
+     specifying the names of the channels this driver is interested to
+     be probed with.
+
+  void unregister_rpmsg_driver(struct rpmsg_driver *rpdrv);
+   - unregisters an rpmsg driver from the rpmsg bus. user should provide
+     a pointer to a previously-registered rpmsg_driver struct.
+     Returns 0 on success, and an appropriate error value on failure.
+
+
+3. Typical usage
+
+The following is a simple rpmsg driver, that sends an "hello!" message
+on probe(), and whenever it receives an incoming message, it dumps its
+content to the console.
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/rpmsg.h>
+
+static void rpmsg_sample_cb(struct rpmsg_channel *rpdev, void *data, int len,
+						void *priv, u32 src)
+{
+	print_hex_dump(KERN_INFO, "incoming message:", DUMP_PREFIX_NONE,
+						16, 1, data, len, true);
+}
+
+static int rpmsg_sample_probe(struct rpmsg_channel *rpdev)
+{
+	int err;
+
+	dev_info(&rpdev->dev, "chnl: 0x%x -> 0x%x\n", rpdev->src, rpdev->dst);
+
+	/* send a message on our channel */
+	err = rpmsg_send(rpdev, "hello!", 6);
+	if (err) {
+		pr_err("rpmsg_send failed: %d\n", err);
+		return err;
+	}
+
+	return 0;
+}
+
+static void __devexit rpmsg_sample_remove(struct rpmsg_channel *rpdev)
+{
+	dev_info(&rpdev->dev, "rpmsg sample client driver is removed\n");
+}
+
+static struct rpmsg_device_id rpmsg_driver_sample_id_table[] = {
+	{ .name	= "rpmsg-client-sample" },
+	{ },
+};
+MODULE_DEVICE_TABLE(rpmsg, rpmsg_driver_sample_id_table);
+
+static struct rpmsg_driver rpmsg_sample_client = {
+	.drv.name	= KBUILD_MODNAME,
+	.drv.owner	= THIS_MODULE,
+	.id_table	= rpmsg_driver_sample_id_table,
+	.probe		= rpmsg_sample_probe,
+	.callback	= rpmsg_sample_cb,
+	.remove		= __devexit_p(rpmsg_sample_remove),
+};
+
+static int __init init(void)
+{
+	return register_rpmsg_driver(&rpmsg_sample_client);
+}
+module_init(init);
+
+static void __exit fini(void)
+{
+	unregister_rpmsg_driver(&rpmsg_sample_client);
+}
+module_exit(fini);
+
+Note: a similar sample which can be built and loaded can be found
+in samples/rpmsg/.
+
+4. Allocations of rpmsg channels:
+
+At this point we only support dynamic allocations of rpmsg channels.
+
+This is possible only with remote processors that have the VIRTIO_RPMSG_F_NS
+virtio device feature set. This feature bit means that the remote
+processor supports dynamic name service announcement messages.
+
+When this feature is enabled, creation of rpmsg devices (i.e. channels)
+is completely dynamic: the remote processor announces the existence of a
+remote rpmsg service by sending a name service message (which contains
+the name and rpmsg addr of the remote service, see struct rpmsg_ns_msg).
+
+This message is then handled by the rpmsg bus, which in turn dynamically
+creates and registers an rpmsg channel (which represents the remote service).
+If/when a relevant rpmsg driver is registered, it will be immediately probed
+by the bus, and can then start sending messages to the remote service.
+
+The plan is also to add static creation of rpmsg channels via the virtio
+config space, but it's not implemented yet.
diff --git a/drivers/Kconfig b/drivers/Kconfig
index 27b34bf41d4c..516faf6d88ba 100644
--- a/drivers/Kconfig
+++ b/drivers/Kconfig
@@ -134,6 +134,8 @@ source "drivers/iommu/Kconfig"
 
 source "drivers/remoteproc/Kconfig"
 
+source "drivers/rpmsg/Kconfig"
+
 source "drivers/virt/Kconfig"
 
 source "drivers/devfreq/Kconfig"
diff --git a/drivers/Makefile b/drivers/Makefile
index f1019b714f27..3fdc17709d36 100644
--- a/drivers/Makefile
+++ b/drivers/Makefile
@@ -127,6 +127,7 @@ obj-$(CONFIG_HWSPINLOCK)	+= hwspinlock/
 obj-$(CONFIG_NFC)		+= nfc/
 obj-$(CONFIG_IOMMU_SUPPORT)	+= iommu/
 obj-$(CONFIG_REMOTEPROC)	+= remoteproc/
+obj-$(CONFIG_RPMSG)		+= rpmsg/
 
 # Virtualization drivers
 obj-$(CONFIG_VIRT_DRIVERS)	+= virt/
diff --git a/drivers/rpmsg/Kconfig b/drivers/rpmsg/Kconfig
new file mode 100644
index 000000000000..811fede35bd8
--- /dev/null
+++ b/drivers/rpmsg/Kconfig
@@ -0,0 +1,5 @@
+# RPMSG always gets selected by whoever wants it
+config RPMSG
+	tristate
+	select VIRTIO
+	select VIRTIO_RING
diff --git a/drivers/rpmsg/Makefile b/drivers/rpmsg/Makefile
new file mode 100644
index 000000000000..7617fcb8259f
--- /dev/null
+++ b/drivers/rpmsg/Makefile
@@ -0,0 +1 @@
+obj-$(CONFIG_RPMSG)	+= virtio_rpmsg_bus.o
diff --git a/drivers/rpmsg/virtio_rpmsg_bus.c b/drivers/rpmsg/virtio_rpmsg_bus.c
new file mode 100644
index 000000000000..257683e7fe8a
--- /dev/null
+++ b/drivers/rpmsg/virtio_rpmsg_bus.c
@@ -0,0 +1,1026 @@
+/*
+ * Virtio-based remote processor messaging bus
+ *
+ * Copyright (C) 2011 Texas Instruments, Inc.
+ * Copyright (C) 2011 Google, Inc.
+ *
+ * Ohad Ben-Cohen <ohad@wizery.com>
+ * Brian Swetland <swetland@google.com>
+ *
+ * This software is licensed under the terms of the GNU General Public
+ * License version 2, as published by the Free Software Foundation, and
+ * may be copied, distributed, and modified under those terms.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#define pr_fmt(fmt) "%s: " fmt, __func__
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/virtio.h>
+#include <linux/virtio_ids.h>
+#include <linux/virtio_config.h>
+#include <linux/scatterlist.h>
+#include <linux/dma-mapping.h>
+#include <linux/slab.h>
+#include <linux/idr.h>
+#include <linux/jiffies.h>
+#include <linux/sched.h>
+#include <linux/wait.h>
+#include <linux/rpmsg.h>
+#include <linux/mutex.h>
+
+/**
+ * struct virtproc_info - virtual remote processor state
+ * @vdev:	the virtio device
+ * @rvq:	rx virtqueue
+ * @svq:	tx virtqueue
+ * @rbufs:	kernel address of rx buffers
+ * @sbufs:	kernel address of tx buffers
+ * @last_sbuf:	index of last tx buffer used
+ * @bufs_dma:	dma base addr of the buffers
+ * @tx_lock:	protects svq, sbufs and sleepers, to allow concurrent senders.
+ *		sending a message might require waking up a dozing remote
+ *		processor, which involves sleeping, hence the mutex.
+ * @endpoints:	idr of local endpoints, allows fast retrieval
+ * @endpoints_lock: lock of the endpoints set
+ * @sendq:	wait queue of sending contexts waiting for a tx buffers
+ * @sleepers:	number of senders that are waiting for a tx buffer
+ * @ns_ept:	the bus's name service endpoint
+ *
+ * This structure stores the rpmsg state of a given virtio remote processor
+ * device (there might be several virtio proc devices for each physical
+ * remote processor).
+ */
+struct virtproc_info {
+	struct virtio_device *vdev;
+	struct virtqueue *rvq, *svq;
+	void *rbufs, *sbufs;
+	int last_sbuf;
+	dma_addr_t bufs_dma;
+	struct mutex tx_lock;
+	struct idr endpoints;
+	struct mutex endpoints_lock;
+	wait_queue_head_t sendq;
+	atomic_t sleepers;
+	struct rpmsg_endpoint *ns_ept;
+};
+
+/**
+ * struct rpmsg_channel_info - internal channel info representation
+ * @name: name of service
+ * @src: local address
+ * @dst: destination address
+ */
+struct rpmsg_channel_info {
+	char name[RPMSG_NAME_SIZE];
+	u32 src;
+	u32 dst;
+};
+
+#define to_rpmsg_channel(d) container_of(d, struct rpmsg_channel, dev)
+#define to_rpmsg_driver(d) container_of(d, struct rpmsg_driver, drv)
+
+/*
+ * We're allocating 512 buffers of 512 bytes for communications, and then
+ * using the first 256 buffers for RX, and the last 256 buffers for TX.
+ *
+ * Each buffer will have 16 bytes for the msg header and 496 bytes for
+ * the payload.
+ *
+ * This will require a total space of 256KB for the buffers.
+ *
+ * We might also want to add support for user-provided buffers in time.
+ * This will allow bigger buffer size flexibility, and can also be used
+ * to achieve zero-copy messaging.
+ *
+ * Note that these numbers are purely a decision of this driver - we
+ * can change this without changing anything in the firmware of the remote
+ * processor.
+ */
+#define RPMSG_NUM_BUFS		(512)
+#define RPMSG_BUF_SIZE		(512)
+#define RPMSG_TOTAL_BUF_SPACE	(RPMSG_NUM_BUFS * RPMSG_BUF_SIZE)
+
+/*
+ * Local addresses are dynamically allocated on-demand.
+ * We do not dynamically assign addresses from the low 1024 range,
+ * in order to reserve that address range for predefined services.
+ */
+#define RPMSG_RESERVED_ADDRESSES	(1024)
+
+/* Address 53 is reserved for advertising remote services */
+#define RPMSG_NS_ADDR			(53)
+
+/* sysfs show configuration fields */
+#define rpmsg_show_attr(field, path, format_string)			\
+static ssize_t								\
+field##_show(struct device *dev,					\
+			struct device_attribute *attr, char *buf)	\
+{									\
+	struct rpmsg_channel *rpdev = to_rpmsg_channel(dev);		\
+									\
+	return sprintf(buf, format_string, rpdev->path);		\
+}
+
+/* for more info, see Documentation/ABI/testing/sysfs-bus-rpmsg */
+rpmsg_show_attr(name, id.name, "%s\n");
+rpmsg_show_attr(src, src, "0x%x\n");
+rpmsg_show_attr(dst, dst, "0x%x\n");
+rpmsg_show_attr(announce, announce ? "true" : "false", "%s\n");
+
+/*
+ * Unique (and free running) index for rpmsg devices.
+ *
+ * Yeah, we're not recycling those numbers (yet?). will be easy
+ * to change if/when we want to.
+ */
+static unsigned int rpmsg_dev_index;
+
+static ssize_t modalias_show(struct device *dev,
+			     struct device_attribute *attr, char *buf)
+{
+	struct rpmsg_channel *rpdev = to_rpmsg_channel(dev);
+
+	return sprintf(buf, RPMSG_DEVICE_MODALIAS_FMT "\n", rpdev->id.name);
+}
+
+static struct device_attribute rpmsg_dev_attrs[] = {
+	__ATTR_RO(name),
+	__ATTR_RO(modalias),
+	__ATTR_RO(dst),
+	__ATTR_RO(src),
+	__ATTR_RO(announce),
+	__ATTR_NULL
+};
+
+/* rpmsg devices and drivers are matched using the service name */
+static inline int rpmsg_id_match(const struct rpmsg_channel *rpdev,
+				  const struct rpmsg_device_id *id)
+{
+	return strncmp(id->name, rpdev->id.name, RPMSG_NAME_SIZE) == 0;
+}
+
+/* match rpmsg channel and rpmsg driver */
+static int rpmsg_dev_match(struct device *dev, struct device_driver *drv)
+{
+	struct rpmsg_channel *rpdev = to_rpmsg_channel(dev);
+	struct rpmsg_driver *rpdrv = to_rpmsg_driver(drv);
+	const struct rpmsg_device_id *ids = rpdrv->id_table;
+	unsigned int i;
+
+	for (i = 0; ids[i].name[0]; i++)
+		if (rpmsg_id_match(rpdev, &ids[i]))
+			return 1;
+
+	return 0;
+}
+
+static int rpmsg_uevent(struct device *dev, struct kobj_uevent_env *env)
+{
+	struct rpmsg_channel *rpdev = to_rpmsg_channel(dev);
+
+	return add_uevent_var(env, "MODALIAS=" RPMSG_DEVICE_MODALIAS_FMT,
+					rpdev->id.name);
+}
+
+/* for more info, see below documentation of rpmsg_create_ept() */
+static struct rpmsg_endpoint *__rpmsg_create_ept(struct virtproc_info *vrp,
+		struct rpmsg_channel *rpdev, rpmsg_rx_cb_t cb,
+		void *priv, u32 addr)
+{
+	int err, tmpaddr, request;
+	struct rpmsg_endpoint *ept;
+	struct device *dev = rpdev ? &rpdev->dev : &vrp->vdev->dev;
+
+	if (!idr_pre_get(&vrp->endpoints, GFP_KERNEL))
+		return NULL;
+
+	ept = kzalloc(sizeof(*ept), GFP_KERNEL);
+	if (!ept) {
+		dev_err(dev, "failed to kzalloc a new ept\n");
+		return NULL;
+	}
+
+	ept->rpdev = rpdev;
+	ept->cb = cb;
+	ept->priv = priv;
+
+	/* do we need to allocate a local address ? */
+	request = addr == RPMSG_ADDR_ANY ? RPMSG_RESERVED_ADDRESSES : addr;
+
+	mutex_lock(&vrp->endpoints_lock);
+
+	/* bind the endpoint to an rpmsg address (and allocate one if needed) */
+	err = idr_get_new_above(&vrp->endpoints, ept, request, &tmpaddr);
+	if (err) {
+		dev_err(dev, "idr_get_new_above failed: %d\n", err);
+		goto free_ept;
+	}
+
+	/* make sure the user's address request is fulfilled, if relevant */
+	if (addr != RPMSG_ADDR_ANY && tmpaddr != addr) {
+		dev_err(dev, "address 0x%x already in use\n", addr);
+		goto rem_idr;
+	}
+
+	ept->addr = tmpaddr;
+
+	mutex_unlock(&vrp->endpoints_lock);
+
+	return ept;
+
+rem_idr:
+	idr_remove(&vrp->endpoints, request);
+free_ept:
+	mutex_unlock(&vrp->endpoints_lock);
+	kfree(ept);
+	return NULL;
+}
+
+/**
+ * rpmsg_create_ept() - create a new rpmsg_endpoint
+ * @rpdev: rpmsg channel device
+ * @cb: rx callback handler
+ * @priv: private data for the driver's use
+ * @addr: local rpmsg address to bind with @cb
+ *
+ * Every rpmsg address in the system is bound to an rx callback (so when
+ * inbound messages arrive, they are dispatched by the rpmsg bus using the
+ * appropriate callback handler) by means of an rpmsg_endpoint struct.
+ *
+ * This function allows drivers to create such an endpoint, and by that,
+ * bind a callback, and possibly some private data too, to an rpmsg address
+ * (either one that is known in advance, or one that will be dynamically
+ * assigned for them).
+ *
+ * Simple rpmsg drivers need not call rpmsg_create_ept, because an endpoint
+ * is already created for them when they are probed by the rpmsg bus
+ * (using the rx callback provided when they registered to the rpmsg bus).
+ *
+ * So things should just work for simple drivers: they already have an
+ * endpoint, their rx callback is bound to their rpmsg address, and when
+ * relevant inbound messages arrive (i.e. messages which their dst address
+ * equals to the src address of their rpmsg channel), the driver's handler
+ * is invoked to process it.
+ *
+ * That said, more complicated drivers might do need to allocate
+ * additional rpmsg addresses, and bind them to different rx callbacks.
+ * To accomplish that, those drivers need to call this function.
+ *
+ * Drivers should provide their @rpdev channel (so the new endpoint would belong
+ * to the same remote processor their channel belongs to), an rx callback
+ * function, an optional private data (which is provided back when the
+ * rx callback is invoked), and an address they want to bind with the
+ * callback. If @addr is RPMSG_ADDR_ANY, then rpmsg_create_ept will
+ * dynamically assign them an available rpmsg address (drivers should have
+ * a very good reason why not to always use RPMSG_ADDR_ANY here).
+ *
+ * Returns a pointer to the endpoint on success, or NULL on error.
+ */
+struct rpmsg_endpoint *rpmsg_create_ept(struct rpmsg_channel *rpdev,
+				rpmsg_rx_cb_t cb, void *priv, u32 addr)
+{
+	return __rpmsg_create_ept(rpdev->vrp, rpdev, cb, priv, addr);
+}
+EXPORT_SYMBOL(rpmsg_create_ept);
+
+/**
+ * rpmsg_destroy_ept() - destroy an existing rpmsg endpoint
+ * @ept: endpoing to destroy
+ *
+ * Should be used by drivers to destroy an rpmsg endpoint previously
+ * created with rpmsg_create_ept().
+ */
+void rpmsg_destroy_ept(struct rpmsg_endpoint *ept)
+{
+	struct virtproc_info *vrp = ept->rpdev->vrp;
+
+	mutex_lock(&vrp->endpoints_lock);
+	idr_remove(&vrp->endpoints, ept->addr);
+	mutex_unlock(&vrp->endpoints_lock);
+
+	kfree(ept);
+}
+EXPORT_SYMBOL(rpmsg_destroy_ept);
+
+/*
+ * when an rpmsg driver is probed with a channel, we seamlessly create
+ * it an endpoint, binding its rx callback to a unique local rpmsg
+ * address.
+ *
+ * if we need to, we also announce about this channel to the remote
+ * processor (needed in case the driver is exposing an rpmsg service).
+ */
+static int rpmsg_dev_probe(struct device *dev)
+{
+	struct rpmsg_channel *rpdev = to_rpmsg_channel(dev);
+	struct rpmsg_driver *rpdrv = to_rpmsg_driver(rpdev->dev.driver);
+	struct virtproc_info *vrp = rpdev->vrp;
+	struct rpmsg_endpoint *ept;
+	int err;
+
+	ept = rpmsg_create_ept(rpdev, rpdrv->callback, NULL, rpdev->src);
+	if (!ept) {
+		dev_err(dev, "failed to create endpoint\n");
+		err = -ENOMEM;
+		goto out;
+	}
+
+	rpdev->ept = ept;
+	rpdev->src = ept->addr;
+
+	err = rpdrv->probe(rpdev);
+	if (err) {
+		dev_err(dev, "%s: failed: %d\n", __func__, err);
+		rpmsg_destroy_ept(ept);
+		goto out;
+	}
+
+	/* need to tell remote processor's name service about this channel ? */
+	if (rpdev->announce &&
+			virtio_has_feature(vrp->vdev, VIRTIO_RPMSG_F_NS)) {
+		struct rpmsg_ns_msg nsm;
+
+		strncpy(nsm.name, rpdev->id.name, RPMSG_NAME_SIZE);
+		nsm.addr = rpdev->src;
+		nsm.flags = RPMSG_NS_CREATE;
+
+		err = rpmsg_sendto(rpdev, &nsm, sizeof(nsm), RPMSG_NS_ADDR);
+		if (err)
+			dev_err(dev, "failed to announce service %d\n", err);
+	}
+
+out:
+	return err;
+}
+
+static int rpmsg_dev_remove(struct device *dev)
+{
+	struct rpmsg_channel *rpdev = to_rpmsg_channel(dev);
+	struct rpmsg_driver *rpdrv = to_rpmsg_driver(rpdev->dev.driver);
+	struct virtproc_info *vrp = rpdev->vrp;
+	int err = 0;
+
+	/* tell remote processor's name service we're removing this channel */
+	if (rpdev->announce &&
+			virtio_has_feature(vrp->vdev, VIRTIO_RPMSG_F_NS)) {
+		struct rpmsg_ns_msg nsm;
+
+		strncpy(nsm.name, rpdev->id.name, RPMSG_NAME_SIZE);
+		nsm.addr = rpdev->src;
+		nsm.flags = RPMSG_NS_DESTROY;
+
+		err = rpmsg_sendto(rpdev, &nsm, sizeof(nsm), RPMSG_NS_ADDR);
+		if (err)
+			dev_err(dev, "failed to announce service %d\n", err);
+	}
+
+	rpdrv->remove(rpdev);
+
+	rpmsg_destroy_ept(rpdev->ept);
+
+	return err;
+}
+
+static struct bus_type rpmsg_bus = {
+	.name		= "rpmsg",
+	.match		= rpmsg_dev_match,
+	.dev_attrs	= rpmsg_dev_attrs,
+	.uevent		= rpmsg_uevent,
+	.probe		= rpmsg_dev_probe,
+	.remove		= rpmsg_dev_remove,
+};
+
+/**
+ * register_rpmsg_driver() - register an rpmsg driver with the rpmsg bus
+ * @rpdrv: pointer to a struct rpmsg_driver
+ *
+ * Returns 0 on success, and an appropriate error value on failure.
+ */
+int register_rpmsg_driver(struct rpmsg_driver *rpdrv)
+{
+	rpdrv->drv.bus = &rpmsg_bus;
+	return driver_register(&rpdrv->drv);
+}
+EXPORT_SYMBOL(register_rpmsg_driver);
+
+/**
+ * unregister_rpmsg_driver() - unregister an rpmsg driver from the rpmsg bus
+ * @rpdrv: pointer to a struct rpmsg_driver
+ *
+ * Returns 0 on success, and an appropriate error value on failure.
+ */
+void unregister_rpmsg_driver(struct rpmsg_driver *rpdrv)
+{
+	driver_unregister(&rpdrv->drv);
+}
+EXPORT_SYMBOL(unregister_rpmsg_driver);
+
+static void rpmsg_release_device(struct device *dev)
+{
+	struct rpmsg_channel *rpdev = to_rpmsg_channel(dev);
+
+	kfree(rpdev);
+}
+
+/*
+ * match an rpmsg channel with a channel info struct.
+ * this is used to make sure we're not creating rpmsg devices for channels
+ * that already exist.
+ */
+static int rpmsg_channel_match(struct device *dev, void *data)
+{
+	struct rpmsg_channel_info *chinfo = data;
+	struct rpmsg_channel *rpdev = to_rpmsg_channel(dev);
+
+	if (chinfo->src != RPMSG_ADDR_ANY && chinfo->src != rpdev->src)
+		return 0;
+
+	if (chinfo->dst != RPMSG_ADDR_ANY && chinfo->dst != rpdev->dst)
+		return 0;
+
+	if (strncmp(chinfo->name, rpdev->id.name, RPMSG_NAME_SIZE))
+		return 0;
+
+	/* found a match ! */
+	return 1;
+}
+
+/*
+ * create an rpmsg channel using its name and address info.
+ * this function will be used to create both static and dynamic
+ * channels.
+ */
+static struct rpmsg_channel *rpmsg_create_channel(struct virtproc_info *vrp,
+				struct rpmsg_channel_info *chinfo)
+{
+	struct rpmsg_channel *rpdev;
+	struct device *tmp, *dev = &vrp->vdev->dev;
+	int ret;
+
+	/* make sure a similar channel doesn't already exist */
+	tmp = device_find_child(dev, chinfo, rpmsg_channel_match);
+	if (tmp) {
+		/* decrement the matched device's refcount back */
+		put_device(tmp);
+		dev_err(dev, "channel %s:%x:%x already exist\n",
+				chinfo->name, chinfo->src, chinfo->dst);
+		return NULL;
+	}
+
+	rpdev = kzalloc(sizeof(struct rpmsg_channel), GFP_KERNEL);
+	if (!rpdev) {
+		pr_err("kzalloc failed\n");
+		return NULL;
+	}
+
+	rpdev->vrp = vrp;
+	rpdev->src = chinfo->src;
+	rpdev->dst = chinfo->dst;
+
+	/*
+	 * rpmsg server channels has predefined local address (for now),
+	 * and their existence needs to be announced remotely
+	 */
+	rpdev->announce = rpdev->src != RPMSG_ADDR_ANY ? true : false;
+
+	strncpy(rpdev->id.name, chinfo->name, RPMSG_NAME_SIZE);
+
+	/* very simple device indexing plumbing which is enough for now */
+	dev_set_name(&rpdev->dev, "rpmsg%d", rpmsg_dev_index++);
+
+	rpdev->dev.parent = &vrp->vdev->dev;
+	rpdev->dev.bus = &rpmsg_bus;
+	rpdev->dev.release = rpmsg_release_device;
+
+	ret = device_register(&rpdev->dev);
+	if (ret) {
+		dev_err(dev, "device_register failed: %d\n", ret);
+		put_device(&rpdev->dev);
+		return NULL;
+	}
+
+	return rpdev;
+}
+
+/*
+ * find an existing channel using its name + address properties,
+ * and destroy it
+ */
+static int rpmsg_destroy_channel(struct virtproc_info *vrp,
+					struct rpmsg_channel_info *chinfo)
+{
+	struct virtio_device *vdev = vrp->vdev;
+	struct device *dev;
+
+	dev = device_find_child(&vdev->dev, chinfo, rpmsg_channel_match);
+	if (!dev)
+		return -EINVAL;
+
+	device_unregister(dev);
+
+	put_device(dev);
+
+	return 0;
+}
+
+/* super simple buffer "allocator" that is just enough for now */
+static void *get_a_tx_buf(struct virtproc_info *vrp)
+{
+	unsigned int len;
+	void *ret;
+
+	/* support multiple concurrent senders */
+	mutex_lock(&vrp->tx_lock);
+
+	/*
+	 * either pick the next unused tx buffer
+	 * (half of our buffers are used for sending messages)
+	 */
+	if (vrp->last_sbuf < RPMSG_NUM_BUFS / 2)
+		ret = vrp->sbufs + RPMSG_BUF_SIZE * vrp->last_sbuf++;
+	/* or recycle a used one */
+	else
+		ret = virtqueue_get_buf(vrp->svq, &len);
+
+	mutex_unlock(&vrp->tx_lock);
+
+	return ret;
+}
+
+/**
+ * rpmsg_upref_sleepers() - enable "tx-complete" interrupts, if needed
+ * @vrp: virtual remote processor state
+ *
+ * This function is called before a sender is blocked, waiting for
+ * a tx buffer to become available.
+ *
+ * If we already have blocking senders, this function merely increases
+ * the "sleepers" reference count, and exits.
+ *
+ * Otherwise, if this is the first sender to block, we also enable
+ * virtio's tx callbacks, so we'd be immediately notified when a tx
+ * buffer is consumed (we rely on virtio's tx callback in order
+ * to wake up sleeping senders as soon as a tx buffer is used by the
+ * remote processor).
+ */
+static void rpmsg_upref_sleepers(struct virtproc_info *vrp)
+{
+	/* support multiple concurrent senders */
+	mutex_lock(&vrp->tx_lock);
+
+	/* are we the first sleeping context waiting for tx buffers ? */
+	if (atomic_inc_return(&vrp->sleepers) == 1)
+		/* enable "tx-complete" interrupts before dozing off */
+		virtqueue_enable_cb(vrp->svq);
+
+	mutex_unlock(&vrp->tx_lock);
+}
+
+/**
+ * rpmsg_downref_sleepers() - disable "tx-complete" interrupts, if needed
+ * @vrp: virtual remote processor state
+ *
+ * This function is called after a sender, that waited for a tx buffer
+ * to become available, is unblocked.
+ *
+ * If we still have blocking senders, this function merely decreases
+ * the "sleepers" reference count, and exits.
+ *
+ * Otherwise, if there are no more blocking senders, we also disable
+ * virtio's tx callbacks, to avoid the overhead incurred with handling
+ * those (now redundant) interrupts.
+ */
+static void rpmsg_downref_sleepers(struct virtproc_info *vrp)
+{
+	/* support multiple concurrent senders */
+	mutex_lock(&vrp->tx_lock);
+
+	/* are we the last sleeping context waiting for tx buffers ? */
+	if (atomic_dec_and_test(&vrp->sleepers))
+		/* disable "tx-complete" interrupts */
+		virtqueue_disable_cb(vrp->svq);
+
+	mutex_unlock(&vrp->tx_lock);
+}
+
+/**
+ * rpmsg_send_offchannel_raw() - send a message across to the remote processor
+ * @rpdev: the rpmsg channel
+ * @src: source address
+ * @dst: destination address
+ * @data: payload of message
+ * @len: length of payload
+ * @wait: indicates whether caller should block in case no TX buffers available
+ *
+ * This function is the base implementation for all of the rpmsg sending API.
+ *
+ * It will send @data of length @len to @dst, and say it's from @src. The
+ * message will be sent to the remote processor which the @rpdev channel
+ * belongs to.
+ *
+ * The message is sent using one of the TX buffers that are available for
+ * communication with this remote processor.
+ *
+ * If @wait is true, the caller will be blocked until either a TX buffer is
+ * available, or 15 seconds elapses (we don't want callers to
+ * sleep indefinitely due to misbehaving remote processors), and in that
+ * case -ERESTARTSYS is returned. The number '15' itself was picked
+ * arbitrarily; there's little point in asking drivers to provide a timeout
+ * value themselves.
+ *
+ * Otherwise, if @wait is false, and there are no TX buffers available,
+ * the function will immediately fail, and -ENOMEM will be returned.
+ *
+ * Normally drivers shouldn't use this function directly; instead, drivers
+ * should use the appropriate rpmsg_{try}send{to, _offchannel} API
+ * (see include/linux/rpmsg.h).
+ *
+ * Returns 0 on success and an appropriate error value on failure.
+ */
+int rpmsg_send_offchannel_raw(struct rpmsg_channel *rpdev, u32 src, u32 dst,
+					void *data, int len, bool wait)
+{
+	struct virtproc_info *vrp = rpdev->vrp;
+	struct device *dev = &rpdev->dev;
+	struct scatterlist sg;
+	struct rpmsg_hdr *msg;
+	int err;
+
+	/* bcasting isn't allowed */
+	if (src == RPMSG_ADDR_ANY || dst == RPMSG_ADDR_ANY) {
+		dev_err(dev, "invalid addr (src 0x%x, dst 0x%x)\n", src, dst);
+		return -EINVAL;
+	}
+
+	/*
+	 * We currently use fixed-sized buffers, and therefore the payload
+	 * length is limited.
+	 *
+	 * One of the possible improvements here is either to support
+	 * user-provided buffers (and then we can also support zero-copy
+	 * messaging), or to improve the buffer allocator, to support
+	 * variable-length buffer sizes.
+	 */
+	if (len > RPMSG_BUF_SIZE - sizeof(struct rpmsg_hdr)) {
+		dev_err(dev, "message is too big (%d)\n", len);
+		return -EMSGSIZE;
+	}
+
+	/* grab a buffer */
+	msg = get_a_tx_buf(vrp);
+	if (!msg && !wait)
+		return -ENOMEM;
+
+	/* no free buffer ? wait for one (but bail after 15 seconds) */
+	while (!msg) {
+		/* enable "tx-complete" interrupts, if not already enabled */
+		rpmsg_upref_sleepers(vrp);
+
+		/*
+		 * sleep until a free buffer is available or 15 secs elapse.
+		 * the timeout period is not configurable because there's
+		 * little point in asking drivers to specify that.
+		 * if later this happens to be required, it'd be easy to add.
+		 */
+		err = wait_event_interruptible_timeout(vrp->sendq,
+					(msg = get_a_tx_buf(vrp)),
+					msecs_to_jiffies(15000));
+
+		/* disable "tx-complete" interrupts if we're the last sleeper */
+		rpmsg_downref_sleepers(vrp);
+
+		/* timeout ? */
+		if (!err) {
+			dev_err(dev, "timeout waiting for a tx buffer\n");
+			return -ERESTARTSYS;
+		}
+	}
+
+	msg->len = len;
+	msg->flags = 0;
+	msg->src = src;
+	msg->dst = dst;
+	msg->reserved = 0;
+	memcpy(msg->data, data, len);
+
+	dev_dbg(dev, "TX From 0x%x, To 0x%x, Len %d, Flags %d, Reserved %d\n",
+					msg->src, msg->dst, msg->len,
+					msg->flags, msg->reserved);
+	print_hex_dump(KERN_DEBUG, "rpmsg_virtio TX: ", DUMP_PREFIX_NONE, 16, 1,
+					msg, sizeof(*msg) + msg->len, true);
+
+	sg_init_one(&sg, msg, sizeof(*msg) + len);
+
+	mutex_lock(&vrp->tx_lock);
+
+	/* add message to the remote processor's virtqueue */
+	err = virtqueue_add_buf_gfp(vrp->svq, &sg, 1, 0, msg, GFP_KERNEL);
+	if (err < 0) {
+		/*
+		 * need to reclaim the buffer here, otherwise it's lost
+		 * (memory won't leak, but rpmsg won't use it again for TX).
+		 * this will wait for a buffer management overhaul.
+		 */
+		dev_err(dev, "virtqueue_add_buf_gfp failed: %d\n", err);
+		goto out;
+	}
+
+	/* tell the remote processor it has a pending message to read */
+	virtqueue_kick(vrp->svq);
+
+	err = 0;
+out:
+	mutex_unlock(&vrp->tx_lock);
+	return err;
+}
+EXPORT_SYMBOL(rpmsg_send_offchannel_raw);
+
+/* called when an rx buffer is used, and it's time to digest a message */
+static void rpmsg_recv_done(struct virtqueue *rvq)
+{
+	struct rpmsg_hdr *msg;
+	unsigned int len;
+	struct rpmsg_endpoint *ept;
+	struct scatterlist sg;
+	struct virtproc_info *vrp = rvq->vdev->priv;
+	struct device *dev = &rvq->vdev->dev;
+	int err;
+
+	msg = virtqueue_get_buf(rvq, &len);
+	if (!msg) {
+		dev_err(dev, "uhm, incoming signal, but no used buffer ?\n");
+		return;
+	}
+
+	dev_dbg(dev, "From: 0x%x, To: 0x%x, Len: %d, Flags: %d, Reserved: %d\n",
+					msg->src, msg->dst, msg->len,
+					msg->flags, msg->reserved);
+	print_hex_dump(KERN_DEBUG, "rpmsg_virtio RX: ", DUMP_PREFIX_NONE, 16, 1,
+					msg, sizeof(*msg) + msg->len, true);
+
+	/* use the dst addr to fetch the callback of the appropriate user */
+	mutex_lock(&vrp->endpoints_lock);
+	ept = idr_find(&vrp->endpoints, msg->dst);
+	mutex_unlock(&vrp->endpoints_lock);
+
+	if (ept && ept->cb)
+		ept->cb(ept->rpdev, msg->data, msg->len, ept->priv, msg->src);
+	else
+		dev_warn(dev, "msg received with no recepient\n");
+
+	sg_init_one(&sg, msg, sizeof(*msg) + len);
+
+	/* add the buffer back to the remote processor's virtqueue */
+	err = virtqueue_add_buf_gfp(vrp->rvq, &sg, 0, 1, msg, GFP_KERNEL);
+	if (err < 0) {
+		dev_err(dev, "failed to add a virtqueue buffer: %d\n", err);
+		return;
+	}
+
+	/* tell the remote processor we added another available rx buffer */
+	virtqueue_kick(vrp->rvq);
+}
+
+/*
+ * This is invoked whenever the remote processor completed processing
+ * a TX msg we just sent it, and the buffer is put back to the used ring.
+ *
+ * Normally, though, we suppress this "tx complete" interrupt in order to
+ * avoid the incurred overhead.
+ */
+static void rpmsg_xmit_done(struct virtqueue *svq)
+{
+	struct virtproc_info *vrp = svq->vdev->priv;
+
+	dev_dbg(&svq->vdev->dev, "%s\n", __func__);
+
+	/* wake up potential senders that are waiting for a tx buffer */
+	wake_up_interruptible(&vrp->sendq);
+}
+
+/* invoked when a name service announcement arrives */
+static void rpmsg_ns_cb(struct rpmsg_channel *rpdev, void *data, int len,
+							void *priv, u32 src)
+{
+	struct rpmsg_ns_msg *msg = data;
+	struct rpmsg_channel *newch;
+	struct rpmsg_channel_info chinfo;
+	struct virtproc_info *vrp = priv;
+	struct device *dev = &vrp->vdev->dev;
+	int ret;
+
+	print_hex_dump(KERN_DEBUG, "NS announcement: ",
+			DUMP_PREFIX_NONE, 16, 1,
+			data, len, true);
+
+	if (len != sizeof(*msg)) {
+		dev_err(dev, "malformed ns msg (%d)\n", len);
+		return;
+	}
+
+	/*
+	 * the name service ept does _not_ belong to a real rpmsg channel,
+	 * and is handled by the rpmsg bus itself.
+	 * for sanity reasons, make sure a valid rpdev has _not_ sneaked
+	 * in somehow.
+	 */
+	if (rpdev) {
+		dev_err(dev, "anomaly: ns ept has an rpdev handle\n");
+		return;
+	}
+
+	/* don't trust the remote processor for null terminating the name */
+	msg->name[RPMSG_NAME_SIZE - 1] = '\0';
+
+	dev_info(dev, "%sing channel %s addr 0x%x\n",
+			msg->flags & RPMSG_NS_DESTROY ? "destroy" : "creat",
+			msg->name, msg->addr);
+
+	strncpy(chinfo.name, msg->name, sizeof(chinfo.name));
+	chinfo.src = RPMSG_ADDR_ANY;
+	chinfo.dst = msg->addr;
+
+	if (msg->flags & RPMSG_NS_DESTROY) {
+		ret = rpmsg_destroy_channel(vrp, &chinfo);
+		if (ret)
+			dev_err(dev, "rpmsg_destroy_channel failed: %d\n", ret);
+	} else {
+		newch = rpmsg_create_channel(vrp, &chinfo);
+		if (!newch)
+			dev_err(dev, "rpmsg_create_channel failed\n");
+	}
+}
+
+static int rpmsg_probe(struct virtio_device *vdev)
+{
+	vq_callback_t *vq_cbs[] = { rpmsg_recv_done, rpmsg_xmit_done };
+	const char *names[] = { "input", "output" };
+	struct virtqueue *vqs[2];
+	struct virtproc_info *vrp;
+	void *bufs_va;
+	int err = 0, i;
+
+	vrp = kzalloc(sizeof(*vrp), GFP_KERNEL);
+	if (!vrp)
+		return -ENOMEM;
+
+	vrp->vdev = vdev;
+
+	idr_init(&vrp->endpoints);
+	mutex_init(&vrp->endpoints_lock);
+	mutex_init(&vrp->tx_lock);
+	init_waitqueue_head(&vrp->sendq);
+
+	/* We expect two virtqueues, rx and tx (and in this order) */
+	err = vdev->config->find_vqs(vdev, 2, vqs, vq_cbs, names);
+	if (err)
+		goto free_vrp;
+
+	vrp->rvq = vqs[0];
+	vrp->svq = vqs[1];
+
+	/* allocate coherent memory for the buffers */
+	bufs_va = dma_alloc_coherent(vdev->dev.parent, RPMSG_TOTAL_BUF_SPACE,
+				&vrp->bufs_dma, GFP_KERNEL);
+	if (!bufs_va)
+		goto vqs_del;
+
+	dev_dbg(&vdev->dev, "buffers: va %p, dma 0x%x\n", bufs_va,
+						vrp->bufs_dma);
+
+	/* half of the buffers is dedicated for RX */
+	vrp->rbufs = bufs_va;
+
+	/* and half is dedicated for TX */
+	vrp->sbufs = bufs_va + RPMSG_TOTAL_BUF_SPACE / 2;
+
+	/* set up the receive buffers */
+	for (i = 0; i < RPMSG_NUM_BUFS / 2; i++) {
+		struct scatterlist sg;
+		void *cpu_addr = vrp->rbufs + i * RPMSG_BUF_SIZE;
+
+		sg_init_one(&sg, cpu_addr, RPMSG_BUF_SIZE);
+
+		err = virtqueue_add_buf_gfp(vrp->rvq, &sg, 0, 1, cpu_addr,
+								GFP_KERNEL);
+		WARN_ON(err < 0); /* sanity check; this can't really happen */
+	}
+
+	/* suppress "tx-complete" interrupts */
+	virtqueue_disable_cb(vrp->svq);
+
+	vdev->priv = vrp;
+
+	/* if supported by the remote processor, enable the name service */
+	if (virtio_has_feature(vdev, VIRTIO_RPMSG_F_NS)) {
+		/* a dedicated endpoint handles the name service msgs */
+		vrp->ns_ept = __rpmsg_create_ept(vrp, NULL, rpmsg_ns_cb,
+						vrp, RPMSG_NS_ADDR);
+		if (!vrp->ns_ept) {
+			dev_err(&vdev->dev, "failed to create the ns ept\n");
+			err = -ENOMEM;
+			goto free_coherent;
+		}
+	}
+
+	/* tell the remote processor it can start sending messages */
+	virtqueue_kick(vrp->rvq);
+
+	dev_info(&vdev->dev, "rpmsg host is online\n");
+
+	return 0;
+
+free_coherent:
+	dma_free_coherent(vdev->dev.parent, RPMSG_TOTAL_BUF_SPACE, bufs_va,
+					vrp->bufs_dma);
+vqs_del:
+	vdev->config->del_vqs(vrp->vdev);
+free_vrp:
+	kfree(vrp);
+	return err;
+}
+
+static int rpmsg_remove_device(struct device *dev, void *data)
+{
+	device_unregister(dev);
+
+	return 0;
+}
+
+static void __devexit rpmsg_remove(struct virtio_device *vdev)
+{
+	struct virtproc_info *vrp = vdev->priv;
+	int ret;
+
+	vdev->config->reset(vdev);
+
+	ret = device_for_each_child(&vdev->dev, NULL, rpmsg_remove_device);
+	if (ret)
+		dev_warn(&vdev->dev, "can't remove rpmsg device: %d\n", ret);
+
+	idr_remove_all(&vrp->endpoints);
+	idr_destroy(&vrp->endpoints);
+
+	vdev->config->del_vqs(vrp->vdev);
+
+	dma_free_coherent(vdev->dev.parent, RPMSG_TOTAL_BUF_SPACE,
+					vrp->rbufs, vrp->bufs_dma);
+
+	kfree(vrp);
+}
+
+static struct virtio_device_id id_table[] = {
+	{ VIRTIO_ID_RPMSG, VIRTIO_DEV_ANY_ID },
+	{ 0 },
+};
+
+static unsigned int features[] = {
+	VIRTIO_RPMSG_F_NS,
+};
+
+static struct virtio_driver virtio_ipc_driver = {
+	.feature_table	= features,
+	.feature_table_size = ARRAY_SIZE(features),
+	.driver.name	= KBUILD_MODNAME,
+	.driver.owner	= THIS_MODULE,
+	.id_table	= id_table,
+	.probe		= rpmsg_probe,
+	.remove		= __devexit_p(rpmsg_remove),
+};
+
+static int __init rpmsg_init(void)
+{
+	int ret;
+
+	ret = bus_register(&rpmsg_bus);
+	if (ret) {
+		pr_err("failed to register rpmsg bus: %d\n", ret);
+		return ret;
+	}
+
+	ret = register_virtio_driver(&virtio_ipc_driver);
+	if (ret) {
+		pr_err("failed to register virtio driver: %d\n", ret);
+		bus_unregister(&rpmsg_bus);
+	}
+
+	return ret;
+}
+module_init(rpmsg_init);
+
+static void __exit rpmsg_fini(void)
+{
+	unregister_virtio_driver(&virtio_ipc_driver);
+	bus_unregister(&rpmsg_bus);
+}
+module_exit(rpmsg_fini);
+
+MODULE_DEVICE_TABLE(virtio, id_table);
+MODULE_DESCRIPTION("Virtio-based remote processor messaging bus");
+MODULE_LICENSE("GPL v2");
diff --git a/include/linux/mod_devicetable.h b/include/linux/mod_devicetable.h
index b29e7f6f8fa5..92aef8aaef1a 100644
--- a/include/linux/mod_devicetable.h
+++ b/include/linux/mod_devicetable.h
@@ -414,6 +414,15 @@ struct hv_vmbus_device_id {
 			__attribute__((aligned(sizeof(kernel_ulong_t))));
 };
 
+/* rpmsg */
+
+#define RPMSG_NAME_SIZE			32
+#define RPMSG_DEVICE_MODALIAS_FMT	"rpmsg:%s"
+
+struct rpmsg_device_id {
+	char name[RPMSG_NAME_SIZE];
+};
+
 /* i2c */
 
 #define I2C_NAME_SIZE	20
diff --git a/include/linux/rpmsg.h b/include/linux/rpmsg.h
new file mode 100644
index 000000000000..a8e50e44203c
--- /dev/null
+++ b/include/linux/rpmsg.h
@@ -0,0 +1,326 @@
+/*
+ * Remote processor messaging
+ *
+ * Copyright (C) 2011 Texas Instruments, Inc.
+ * Copyright (C) 2011 Google, Inc.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ *   notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ *   notice, this list of conditions and the following disclaimer in
+ *   the documentation and/or other materials provided with the
+ *   distribution.
+ * * Neither the name Texas Instruments nor the names of its
+ *   contributors may be used to endorse or promote products derived
+ *   from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _LINUX_RPMSG_H
+#define _LINUX_RPMSG_H
+
+#include <linux/types.h>
+#include <linux/device.h>
+#include <linux/mod_devicetable.h>
+
+/* The feature bitmap for virtio rpmsg */
+#define VIRTIO_RPMSG_F_NS	0 /* RP supports name service notifications */
+
+/**
+ * struct rpmsg_hdr - common header for all rpmsg messages
+ * @src: source address
+ * @dst: destination address
+ * @reserved: reserved for future use
+ * @len: length of payload (in bytes)
+ * @flags: message flags
+ * @data: @len bytes of message payload data
+ *
+ * Every message sent(/received) on the rpmsg bus begins with this header.
+ */
+struct rpmsg_hdr {
+	u32 src;
+	u32 dst;
+	u32 reserved;
+	u16 len;
+	u16 flags;
+	u8 data[0];
+} __packed;
+
+/**
+ * struct rpmsg_ns_msg - dynamic name service announcement message
+ * @name: name of remote service that is published
+ * @addr: address of remote service that is published
+ * @flags: indicates whether service is created or destroyed
+ *
+ * This message is sent across to publish a new service, or announce
+ * about its removal. When we receive these messages, an appropriate
+ * rpmsg channel (i.e device) is created/destroyed. In turn, the ->probe()
+ * or ->remove() handler of the appropriate rpmsg driver will be invoked
+ * (if/as-soon-as one is registered).
+ */
+struct rpmsg_ns_msg {
+	char name[RPMSG_NAME_SIZE];
+	u32 addr;
+	u32 flags;
+} __packed;
+
+/**
+ * enum rpmsg_ns_flags - dynamic name service announcement flags
+ *
+ * @RPMSG_NS_CREATE: a new remote service was just created
+ * @RPMSG_NS_DESTROY: a known remote service was just destroyed
+ */
+enum rpmsg_ns_flags {
+	RPMSG_NS_CREATE		= 0,
+	RPMSG_NS_DESTROY	= 1,
+};
+
+#define RPMSG_ADDR_ANY		0xFFFFFFFF
+
+struct virtproc_info;
+
+/**
+ * rpmsg_channel - devices that belong to the rpmsg bus are called channels
+ * @vrp: the remote processor this channel belongs to
+ * @dev: the device struct
+ * @id: device id (used to match between rpmsg drivers and devices)
+ * @src: local address
+ * @dst: destination address
+ * @ept: the rpmsg endpoint of this channel
+ * @announce: if set, rpmsg will announce the creation/removal of this channel
+ */
+struct rpmsg_channel {
+	struct virtproc_info *vrp;
+	struct device dev;
+	struct rpmsg_device_id id;
+	u32 src;
+	u32 dst;
+	struct rpmsg_endpoint *ept;
+	bool announce;
+};
+
+typedef void (*rpmsg_rx_cb_t)(struct rpmsg_channel *, void *, int, void *, u32);
+
+/**
+ * struct rpmsg_endpoint - binds a local rpmsg address to its user
+ * @rpdev: rpmsg channel device
+ * @cb: rx callback handler
+ * @addr: local rpmsg address
+ * @priv: private data for the driver's use
+ *
+ * In essence, an rpmsg endpoint represents a listener on the rpmsg bus, as
+ * it binds an rpmsg address with an rx callback handler.
+ *
+ * Simple rpmsg drivers shouldn't use this struct directly, because
+ * things just work: every rpmsg driver provides an rx callback upon
+ * registering to the bus, and that callback is then bound to its rpmsg
+ * address when the driver is probed. When relevant inbound messages arrive
+ * (i.e. messages which their dst address equals to the src address of
+ * the rpmsg channel), the driver's handler is invoked to process it.
+ *
+ * More complicated drivers though, that do need to allocate additional rpmsg
+ * addresses, and bind them to different rx callbacks, must explicitly
+ * create additional endpoints by themselves (see rpmsg_create_ept()).
+ */
+struct rpmsg_endpoint {
+	struct rpmsg_channel *rpdev;
+	rpmsg_rx_cb_t cb;
+	u32 addr;
+	void *priv;
+};
+
+/**
+ * struct rpmsg_driver - rpmsg driver struct
+ * @drv: underlying device driver
+ * @id_table: rpmsg ids serviced by this driver
+ * @probe: invoked when a matching rpmsg channel (i.e. device) is found
+ * @remove: invoked when the rpmsg channel is removed
+ * @callback: invoked when an inbound message is received on the channel
+ */
+struct rpmsg_driver {
+	struct device_driver drv;
+	const struct rpmsg_device_id *id_table;
+	int (*probe)(struct rpmsg_channel *dev);
+	void (*remove)(struct rpmsg_channel *dev);
+	void (*callback)(struct rpmsg_channel *, void *, int, void *, u32);
+};
+
+int register_rpmsg_device(struct rpmsg_channel *dev);
+void unregister_rpmsg_device(struct rpmsg_channel *dev);
+int register_rpmsg_driver(struct rpmsg_driver *drv);
+void unregister_rpmsg_driver(struct rpmsg_driver *drv);
+void rpmsg_destroy_ept(struct rpmsg_endpoint *);
+struct rpmsg_endpoint *rpmsg_create_ept(struct rpmsg_channel *,
+				rpmsg_rx_cb_t cb, void *priv, u32 addr);
+int
+rpmsg_send_offchannel_raw(struct rpmsg_channel *, u32, u32, void *, int, bool);
+
+/**
+ * rpmsg_send() - send a message across to the remote processor
+ * @rpdev: the rpmsg channel
+ * @data: payload of message
+ * @len: length of payload
+ *
+ * This function sends @data of length @len on the @rpdev channel.
+ * The message will be sent to the remote processor which the @rpdev
+ * channel belongs to, using @rpdev's source and destination addresses.
+ * In case there are no TX buffers available, the function will block until
+ * one becomes available, or a timeout of 15 seconds elapses. When the latter
+ * happens, -ERESTARTSYS is returned.
+ *
+ * Can only be called from process context (for now).
+ *
+ * Returns 0 on success and an appropriate error value on failure.
+ */
+static inline int rpmsg_send(struct rpmsg_channel *rpdev, void *data, int len)
+{
+	u32 src = rpdev->src, dst = rpdev->dst;
+
+	return rpmsg_send_offchannel_raw(rpdev, src, dst, data, len, true);
+}
+
+/**
+ * rpmsg_sendto() - send a message across to the remote processor, specify dst
+ * @rpdev: the rpmsg channel
+ * @data: payload of message
+ * @len: length of payload
+ * @dst: destination address
+ *
+ * This function sends @data of length @len to the remote @dst address.
+ * The message will be sent to the remote processor which the @rpdev
+ * channel belongs to, using @rpdev's source address.
+ * In case there are no TX buffers available, the function will block until
+ * one becomes available, or a timeout of 15 seconds elapses. When the latter
+ * happens, -ERESTARTSYS is returned.
+ *
+ * Can only be called from process context (for now).
+ *
+ * Returns 0 on success and an appropriate error value on failure.
+ */
+static inline
+int rpmsg_sendto(struct rpmsg_channel *rpdev, void *data, int len, u32 dst)
+{
+	u32 src = rpdev->src;
+
+	return rpmsg_send_offchannel_raw(rpdev, src, dst, data, len, true);
+}
+
+/**
+ * rpmsg_send_offchannel() - send a message using explicit src/dst addresses
+ * @rpdev: the rpmsg channel
+ * @src: source address
+ * @dst: destination address
+ * @data: payload of message
+ * @len: length of payload
+ *
+ * This function sends @data of length @len to the remote @dst address,
+ * and uses @src as the source address.
+ * The message will be sent to the remote processor which the @rpdev
+ * channel belongs to.
+ * In case there are no TX buffers available, the function will block until
+ * one becomes available, or a timeout of 15 seconds elapses. When the latter
+ * happens, -ERESTARTSYS is returned.
+ *
+ * Can only be called from process context (for now).
+ *
+ * Returns 0 on success and an appropriate error value on failure.
+ */
+static inline
+int rpmsg_send_offchannel(struct rpmsg_channel *rpdev, u32 src, u32 dst,
+							void *data, int len)
+{
+	return rpmsg_send_offchannel_raw(rpdev, src, dst, data, len, true);
+}
+
+/**
+ * rpmsg_send() - send a message across to the remote processor
+ * @rpdev: the rpmsg channel
+ * @data: payload of message
+ * @len: length of payload
+ *
+ * This function sends @data of length @len on the @rpdev channel.
+ * The message will be sent to the remote processor which the @rpdev
+ * channel belongs to, using @rpdev's source and destination addresses.
+ * In case there are no TX buffers available, the function will immediately
+ * return -ENOMEM without waiting until one becomes available.
+ *
+ * Can only be called from process context (for now).
+ *
+ * Returns 0 on success and an appropriate error value on failure.
+ */
+static inline
+int rpmsg_trysend(struct rpmsg_channel *rpdev, void *data, int len)
+{
+	u32 src = rpdev->src, dst = rpdev->dst;
+
+	return rpmsg_send_offchannel_raw(rpdev, src, dst, data, len, false);
+}
+
+/**
+ * rpmsg_sendto() - send a message across to the remote processor, specify dst
+ * @rpdev: the rpmsg channel
+ * @data: payload of message
+ * @len: length of payload
+ * @dst: destination address
+ *
+ * This function sends @data of length @len to the remote @dst address.
+ * The message will be sent to the remote processor which the @rpdev
+ * channel belongs to, using @rpdev's source address.
+ * In case there are no TX buffers available, the function will immediately
+ * return -ENOMEM without waiting until one becomes available.
+ *
+ * Can only be called from process context (for now).
+ *
+ * Returns 0 on success and an appropriate error value on failure.
+ */
+static inline
+int rpmsg_trysendto(struct rpmsg_channel *rpdev, void *data, int len, u32 dst)
+{
+	u32 src = rpdev->src;
+
+	return rpmsg_send_offchannel_raw(rpdev, src, dst, data, len, false);
+}
+
+/**
+ * rpmsg_send_offchannel() - send a message using explicit src/dst addresses
+ * @rpdev: the rpmsg channel
+ * @src: source address
+ * @dst: destination address
+ * @data: payload of message
+ * @len: length of payload
+ *
+ * This function sends @data of length @len to the remote @dst address,
+ * and uses @src as the source address.
+ * The message will be sent to the remote processor which the @rpdev
+ * channel belongs to.
+ * In case there are no TX buffers available, the function will immediately
+ * return -ENOMEM without waiting until one becomes available.
+ *
+ * Can only be called from process context (for now).
+ *
+ * Returns 0 on success and an appropriate error value on failure.
+ */
+static inline
+int rpmsg_trysend_offchannel(struct rpmsg_channel *rpdev, u32 src, u32 dst,
+							void *data, int len)
+{
+	return rpmsg_send_offchannel_raw(rpdev, src, dst, data, len, false);
+}
+
+#endif /* _LINUX_RPMSG_H */
diff --git a/include/linux/virtio_ids.h b/include/linux/virtio_ids.h
index 85bb0bb66ffc..b37c5212265e 100644
--- a/include/linux/virtio_ids.h
+++ b/include/linux/virtio_ids.h
@@ -34,6 +34,7 @@
 #define VIRTIO_ID_CONSOLE	3 /* virtio console */
 #define VIRTIO_ID_RNG		4 /* virtio ring */
 #define VIRTIO_ID_BALLOON	5 /* virtio balloon */
+#define VIRTIO_ID_RPMSG		7 /* virtio remote processor messaging */
 #define VIRTIO_ID_9P		9 /* 9p virtio console */
 
 #endif /* _LINUX_VIRTIO_IDS_H */