* Shared memory between remote processors
@ 2017-04-10 9:00 Michele Rodolfi
2017-04-10 19:25 ` Bjorn Andersson
0 siblings, 1 reply; 2+ messages in thread
From: Michele Rodolfi @ 2017-04-10 9:00 UTC (permalink / raw)
To: linux-remoteproc
Hi all,
I'm trying to set up a shared memory region for image transfers
between processors.
Since the rpmsg buffer isn't large enough to contain a whole image, I
would like to put the image into a shared memory buffer then send the
physical address of this buffer to the remote processor using rpmsg.
I'm trying to understand how to do so with the functionalities
provided by remoteproc. Has anyone done this before?
The remoteproc driver supports the request for allocation of a
physically contiguous memory region via resource table. Could this be
a solution?
Thanks
Michele
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: Shared memory between remote processors
2017-04-10 9:00 Shared memory between remote processors Michele Rodolfi
@ 2017-04-10 19:25 ` Bjorn Andersson
0 siblings, 0 replies; 2+ messages in thread
From: Bjorn Andersson @ 2017-04-10 19:25 UTC (permalink / raw)
To: Michele Rodolfi; +Cc: linux-remoteproc
On Mon 10 Apr 02:00 PDT 2017, Michele Rodolfi wrote:
> Hi all,
> I'm trying to set up a shared memory region for image transfers
> between processors.
> Since the rpmsg buffer isn't large enough to contain a whole image, I
> would like to put the image into a shared memory buffer then send the
> physical address of this buffer to the remote processor using rpmsg.
> I'm trying to understand how to do so with the functionalities
> provided by remoteproc. Has anyone done this before?
> The remoteproc driver supports the request for allocation of a
> physically contiguous memory region via resource table. Could this be
> a solution?
>
Specifying a carveout region in your resource table and use that for
your shared buffers sounds plausable, but today we do not have a
mechanism for a rpmsg client driver to acquire a handle to an arbitrary
carveout from a potential parenting remoteproc instance.
This is something that has been discussed, but I have not seem any good
suggestions on how to link the two pieces together.
But as it sounds like you're going to have a "control protocol" running
ontop of rpmsg to provide the remote firmware with buffers to operate on
I would suggest that you allocate the buffers (e.g. using
dma_alloc_coherent()) in the context of that implementation, rather than
tying them to remoteproc.
Regards,
Bjorn
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2017-04-10 19:25 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2017-04-10 9:00 Shared memory between remote processors Michele Rodolfi
2017-04-10 19:25 ` Bjorn Andersson
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox