When inbound messages arrive, rpmsg core looks up their associated
endpoint (by destination address) and then invokes their callback.
We've made sure that endpoints will never be de-allocated after they
were found by rpmsg core, but we also need to protect against the
(rare) scenario where the rpmsg driver was just removed, and its
callback function isn't available anymore.
This is achieved by introducing a callback mutex, which must be taken
before the callback is invoked, and, obviously, before it is removed.
Cc: stable <stable@vger.kernel.org>
Reported-by: Fernando Guzman Lugo <fernando.lugo@ti.com>
Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
When an inbound message arrives, the rpmsg core looks up its
associated endpoint and invokes the registered callback.
If a message arrives while its endpoint is being removed (because
the rpmsg driver was removed, or a recovery of a remote processor
has kicked in) we must ensure atomicity, i.e.:
- Either the ept is removed before it is found
or
- The ept is found but will not be freed until the callback returns
This is achieved by maintaining a per-ept reference count, which,
when drops to zero, will trigger deallocation of the ept.
With this in hand, it is now forbidden to directly deallocate
epts once they have been added to the endpoints idr.
Cc: stable <stable@vger.kernel.org>
Reported-by: Fernando Guzman Lugo <fernando.lugo@ti.com>
Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
dev_dbg() in rpmsg_probe() made use of the %x formatting that
expects an 'unsigned int' which dma_addr_t is not in cases where
dma_addr_t is 64-bit (CONFIG_ARCH_DMA_ADDR_T_64BIT). Casting to
a 'unsigned long long' and using %llx will avoid this.
Signed-off-by: Mark Asselstine <mark.asselstine@windriver.com>
CC: Ohad Ben-Cohen <ohad@wizery.com>
CC: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
After processing an incoming message, always publish the real size
of its containing buffer when putting it back on the available rx ring.
Using any different value might erroneously limit the remote processor
(leading it to think the buffer is smaller than it really is).
Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Cc: Grant Likely <grant.likely@secretlab.ca>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Mark Grosen <mgrosen@ti.com>
Cc: Suman Anna <s-anna@ti.com>
Cc: Fernando Guzman Lugo <fernando.lugo@ti.com>
Cc: Rob Clark <rob@ti.com>
Cc: Ludovic BARRE <ludovic.barre@stericsson.com>
Cc: Loic PALLARDY <loic.pallardy@stericsson.com>
Cc: Omar Ramirez Luna <omar.luna@linaro.org>
When an inbound message arrives, validate its reported length before
propagating it, otherwise buggy (or malicious) remote processors might
trick us into accessing memory which we really shouldn't.
Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Cc: Grant Likely <grant.likely@secretlab.ca>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Mark Grosen <mgrosen@ti.com>
Cc: Suman Anna <s-anna@ti.com>
Cc: Fernando Guzman Lugo <fernando.lugo@ti.com>
Cc: Rob Clark <rob@ti.com>
Cc: Ludovic BARRE <ludovic.barre@stericsson.com>
Cc: Loic PALLARDY <loic.pallardy@stericsson.com>
Cc: Omar Ramirez Luna <omar.luna@linaro.org>
The name service endpoint wasn't destroyed, so fix it.
This is achieved by introducing an internal __rpmsg_destroy_ept
function which doesn't assume the given ept is bound to an rpmsg
channel (much like the existing __rpmsg_create_ept).
This is needed because the name service ept belongs to the rpmsg bus,
and is never bound with a specific rpdev.
Reported-by: Omar Ramirez Luna <omar.ramirez@ti.com>
Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Cc: Grant Likely <grant.likely@secretlab.ca>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Mark Grosen <mgrosen@ti.com>
Cc: Suman Anna <s-anna@ti.com>
Cc: Fernando Guzman Lugo <fernando.lugo@ti.com>
Cc: Rob Clark <rob@ti.com>
Cc: Ludovic BARRE <ludovic.barre@stericsson.com>
Cc: Loic PALLARDY <loic.pallardy@stericsson.com>
Cc: Omar Ramirez Luna <omar.ramirez@ti.com>
Since commit 7bb7aef2 "virtio: rename virtqueue_add_buf_gfp to virtqueue_add_buf",
virtqueue_add_buf_gfp is already rename to virtqueue_add_buf now.
This patch fixes below build error:
CC [M] drivers/rpmsg/virtio_rpmsg_bus.o
drivers/rpmsg/virtio_rpmsg_bus.c: In function 'rpmsg_send_offchannel_raw':
drivers/rpmsg/virtio_rpmsg_bus.c:723: error: implicit declaration of function 'virtqueue_add_buf_gfp'
make[2]: *** [drivers/rpmsg/virtio_rpmsg_bus.o] Error 1
make[1]: *** [drivers/rpmsg] Error 2
make: *** [drivers] Error 2
Signed-off-by: Axel Lin <axel.lin@gmail.com>
Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Add a virtio-based inter-processor communication bus, which enables
kernel drivers to communicate with entities, running on remote
processors, over shared memory using a simple messaging protocol.
Every pair of AMP processors share two vrings, which are used to send
and receive the messages over shared memory.
The header of every message sent on the rpmsg bus contains src and dst
addresses, which make it possible to multiplex several rpmsg channels on
the same vring.
Every rpmsg channel is a device on this bus. When a channel is added,
and an appropriate rpmsg driver is found and probed, it is also assigned
a local rpmsg address, which is then bound to the driver's callback.
When inbound messages carry the local address of a bound driver,
its callback is invoked by the bus.
This patch provides a kernel interface only; user space interfaces
will be later exposed by kernel users of this rpmsg bus.
Designed with Brian Swetland <swetland@google.com>.
Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Acked-by: Rusty Russell <rusty@rustcorp.com.au> (virtio_ids.h)
Cc: Brian Swetland <swetland@google.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Grant Likely <grant.likely@secretlab.ca>
Cc: Tony Lindgren <tony@atomide.com>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Greg KH <greg@kroah.com>
Cc: Stephen Boyd <sboyd@codeaurora.org>
Ohad Ben-Cohen