On Thu, 5 Mar 2020 at 11:07, <nikita.shubin@xxxxxxxxxxx> wrote:
>
>
>
> 05.03.2020, 20:54, "Mathieu Poirier" <mathieu.poirier@xxxxxxxxxx>:
> > On Thu, 5 Mar 2020 at 10:29, <nikita.shubin@xxxxxxxxxxx> wrote:
> >> 05.03.2020, 19:17, "Mathieu Poirier" <mathieu.poirier@xxxxxxxxxx>:
> >> > On Wed, 4 Mar 2020 at 07:25, Nikita Shubin <NShubin@xxxxxxxxxx> wrote:
> >> >> add kick method that does nothing, to avoid errors in rproc_virtio_notify.
> >> >>
> >> >> Signed-off-by: Nikita Shubin <NShubin@xxxxxxxxxx>
> >> >> ---
> >> >> drivers/remoteproc/imx_rproc.c | 6 ++++++
> >> >> 1 file changed, 6 insertions(+)
> >> >>
> >> >> diff --git a/drivers/remoteproc/imx_rproc.c b/drivers/remoteproc/imx_rproc.c
> >> >> index 3e72b6f38d4b..796b6b86550a 100644
> >> >> --- a/drivers/remoteproc/imx_rproc.c
> >> >> +++ b/drivers/remoteproc/imx_rproc.c
> >> >> @@ -240,9 +240,15 @@ static void *imx_rproc_da_to_va(struct rproc *rproc, u64 da, int len)
> >> >> return va;
> >> >> }
> >> >>
> >> >> +static void imx_rproc_kick(struct rproc *rproc, int vqid)
> >> >> +{
> >> >> +
> >> >> +}
> >> >> +
> >> >
> >> > If rproc::kick() is empty, how does the MCU know there is packets to
> >> > fetch in the virtio queues?
> >>
> >> Well, of course it doesn't i understand this perfectly - just following documentation citing:
> >>
> >> | Every remoteproc implementation should at least provide the ->start and ->stop
> >> | handlers. If rpmsg/virtio functionality is also desired, then the ->kick handler
> >> | should be provided as well.
> >>
> >> But i as i mentioned in "remoteproc: Fix NULL pointer dereference in rproc_virtio_notify" kick method will be called if
> >> "resource_table exists in firmware and has "Virtio device entry" defined" anyway, the imx_rproc is not in control of what
> >> exactly it is booting, so such situation can occur.
> >
> > If I understand correctly, the MCU can boot images that have a virtio
> > device in its resource table and still do useful work even if the
> > virtio device/rpmsg bus can't be setup - is this correct?
>
> Yes, this assumption is correct.
>
> Despite this situation is not i desire at all - such thing can happen.
> I am currently using co-proc as a realtime part of UGV control,
> so it must immediately stop the engines, if not provided with navigational information.
>
> The imx7d MCU have access to the most periphery that have the main processor.
>
> Of course the kick method should do real work, but i decided to submit step by step if i am allowed to do so.
Ok, the situation is clearer now and I have put your patches back in
my queue. I am seriously back-logged at this time so it will take a
little while before I get to them.
>
> >
> > Thanks,
> > Mathieu
> >
> >> >
> >> >> static const struct rproc_ops imx_rproc_ops = {
> >> >> .start = imx_rproc_start,
> >> >> .stop = imx_rproc_stop,
> >> >> + .kick = imx_rproc_kick,
> >> >> .da_to_va = imx_rproc_da_to_va,
> >> >> };
> >> >>
> >> >> --
> >> >> 2.24.1
