Re: [PATCH] remoteproc: core: Invoke subdev callbacks in list order

[Bjorn Andersson]

On Tue 25 May 20:16 CDT 2021, Siddharth Gupta wrote:

> 
> On 5/25/2021 5:37 PM, Bjorn Andersson wrote:
> > On Tue 25 May 14:48 CDT 2021, Siddharth Gupta wrote:
> > 
> > > On 5/24/2021 8:03 PM, Bjorn Andersson wrote:
> > > > On Mon 17 May 18:08 CDT 2021, Siddharth Gupta wrote:
> > > > 
> > > > > Subdevices at the beginning of the subdev list should have
> > > > > higher priority than those at the end of the list. Reverse
> > > > > traversal of the list causes priority inversion, which can
> > > > > impact the performance of the device.
> > > > > 
> > > > The subdev lists layers of the communication onion, we bring them up
> > > > inside out and we take them down outside in.
> > > > 
> > > > This stems from the primary idea that we want to be able to shut things
> > > > down cleanly (in the case of a stop) and we pass the "crashed" flag to
> > > > indicate to each recipient during "stop" that it may not rely on the
> > > > response of a lower layer.
> > > > 
> > > > As such, I don't think it's right to say that we have a priority
> > > > inversion.
> > > My understanding of the topic was that each subdevice should be
> > > independent of the other. In our case unfortunately the sysmon
> > > subdevice depends on the glink endpoint.
> > > 
> > We need to care for the ordering if sysmon is to be able to use smd or
> > glink to send the shutdown request.
> Right, I meant the dependence of either sysmon or SSR is on QMI,
> which in turn depends on glink.

The difference between the two is that sysmon ensures that it won't send
any messages to the stopping remoteproc.

> > 
> > > However the priority inversion doesn't happen in these
> > > subdevices, it happens due to the SSR notifications that we send
> > > to kernel clients. In this case kernel clients also can have QMI
> > > sockets that in turn depend on the glink endpoint, which means
> > > when they go to release the QMI socket a broadcast will be sent
> > > out to all connected clients about the closure of the connection
> > > which in this case happens to be the remoteproc which died. So
> > > if we peel the onion, we will be unnecessarily be waiting for a
> > > dead remoteproc.
> > I see, that is indeed a problem.
> > 
> > > > > For example a device adds the glink, sysmon and ssr subdevs
> > > > > to its list. During a crash the ssr notification would go
> > > > > before the glink and sysmon notifications. This can cause a
> > > > > degraded response when a client driver waits for a response
> > > > > from the crashed rproc.
> > > > > 
> > > > In general the design is such that components are not expected to
> > > > communicate with the crashed remote when "crashed" is set, this avoids
> > > > the single-remote crash.
> > > Here the glink device on the rpmsg bus won't know about the
> > > crashed remoteproc till we send glink notification first, right?
> > > Since we send out sysmon and SSR notifications first, the glink
> > > device will still be "alive" on the rpmsg bus.
> > Yes, and this all stems from the design that everything communicating
> > over glink is a child of glink, which isn't the case when you have a SSR
> > event that will end up blocking the sequence in qrtr.
> > 
> > For sysmon this is not a problem, because sysmon is implemented to not
> > attempt to communicate with the parent remoteproc upon a crash.
> Yes, exactly.
> > And all rpmsg devices will be torn down as a result of glink being torn
> > down, so glink can fail early based on this (not sure if this was
> > implemented downstream though).
> This was implemented downstream as a part of an early
> notification that was sent out to the glink device.
> > 
> > > > The case where this isn't holding up is when two remote processors
> > > > crashes simultaneously, in which case e.g. sysmon has been seen hitting
> > > > its timeout waiting for an ack from a dead remoteproc - but I was under
> > > > the impression that this window shrunk dramatically as a side effect of
> > > > us fixing the notification ordering.
> > > You are right, the window would become smaller in the case of two
> > > remoteprocs, but this issue can come up with even a single
> > > remoteproc unless prioritize certain subdevices.
> > The problem that you describe where an SSR notification will directly or
> > indirectly attempt to communicate over QRTR will certainly cause issues
> > in the single-rproc case as well.
> > 
> > 
> > But is there any reason why these listeners has to do the wrong thing at
> > stop(crashed=true)?
> I don't think the listeners are doing anything wrong by closing
> the QMI handle/QRTR socket, the issue is that the glink device
> still thinks that it can communicate.
> > 

The design is such that subdev notifications are not allowed to
communicate with the dying remoteproc when crashed=true.

This means that any listeners to these notifications needs to ensure
they play nice with regards to the dying remoteproc, which includes
sending messages to it or waiting for any new or pending incoming
notifications - QMI or otherwise.

As such I think it makes sense to consider qmi_handle_release() to be an
operation that will send messages and should not be performed in the
notification handler.

> > > > > Signed-off-by: Siddharth Gupta <sidgup@xxxxxxxxxxxxxx>
> > > > > ---
> > > > >    drivers/remoteproc/remoteproc_core.c | 24 ++++++++++++++----------
> > > > >    1 file changed, 14 insertions(+), 10 deletions(-)
> > > > > 
> > > > > diff --git a/drivers/remoteproc/remoteproc_core.c b/drivers/remoteproc/remoteproc_core.c
> > > > > index 626a6b90f..ac8fc42 100644
> > > > > --- a/drivers/remoteproc/remoteproc_core.c
> > > > > +++ b/drivers/remoteproc/remoteproc_core.c
> > > > > @@ -1167,7 +1167,7 @@ static int rproc_handle_resources(struct rproc *rproc,
> > > > >    static int rproc_prepare_subdevices(struct rproc *rproc)
> > > > >    {
> > > > > -	struct rproc_subdev *subdev;
> > > > > +	struct rproc_subdev *subdev, *itr;
> > > > >    	int ret;
> > > > >    	list_for_each_entry(subdev, &rproc->subdevs, node) {
> > > > > @@ -1181,9 +1181,11 @@ static int rproc_prepare_subdevices(struct rproc *rproc)
> > > > >    	return 0;
> > > > >    unroll_preparation:
> > > > > -	list_for_each_entry_continue_reverse(subdev, &rproc->subdevs, node) {
> > > > > -		if (subdev->unprepare)
> > > > > -			subdev->unprepare(subdev);
> > > > > +	list_for_each_entry(itr, &rproc->subdevs, node) {
> > > > > +		if (itr == subdev)
> > > > > +			break;
> > > > > +		if (itr->unprepare)
> > > > > +			itr->unprepare(subdev);
> > > > >    	}
> > > > >    	return ret;
> > > > > @@ -1191,7 +1193,7 @@ static int rproc_prepare_subdevices(struct rproc *rproc)
> > > > >    static int rproc_start_subdevices(struct rproc *rproc)
> > > > >    {
> > > > > -	struct rproc_subdev *subdev;
> > > > > +	struct rproc_subdev *subdev, *itr;
> > > > >    	int ret;
> > > > >    	list_for_each_entry(subdev, &rproc->subdevs, node) {
> > > > > @@ -1205,9 +1207,11 @@ static int rproc_start_subdevices(struct rproc *rproc)
> > > > >    	return 0;
> > > > >    unroll_registration:
> > > > > -	list_for_each_entry_continue_reverse(subdev, &rproc->subdevs, node) {
> > > > > -		if (subdev->stop)
> > > > > -			subdev->stop(subdev, true);
> > > > > +	list_for_each_entry(itr, &rproc->subdevs, node) {
> > > > > +		if (itr == subdev)
> > > > > +			break;
> > > > > +		if (itr->stop)
> > > > > +			itr->stop(itr, true);
> > > > >    	}
> > > > >    	return ret;
> > > > > @@ -1217,7 +1221,7 @@ static void rproc_stop_subdevices(struct rproc *rproc, bool crashed)
> > > > >    {
> > > > >    	struct rproc_subdev *subdev;
> > > > > -	list_for_each_entry_reverse(subdev, &rproc->subdevs, node) {
> > > > > +	list_for_each_entry(subdev, &rproc->subdevs, node) {
> > > > I presume this is the case you actually care about, can you help me
> > > > understand if you changed the others for consistence or if there's some
> > > > flow of events where that might be necessary.
> > > Yes you are right, I only changed the others for consistence.
> > > However, I will give this more thought and see if unprepare in
> > > the reverse order can make a difference.
> > > 
> > Per above argument I don't think things depend on the unrolling on error
> > happening in reverse order. But it's idiomatic.
> I say unprepare in any order might not make a difference because
> prepare would indicate to the subdevice that it should get its
> resources initialized because the remoteproc is going to come up,
> so unprepare would only be the subdevice releasing its resources.
> However start and stop in the reverse order will make a big
> difference. Please correct me if I am wrong.
> 

I think you're right, for the operations we describe today there
shouldn't be any dependencies between the layers for prepare and
unprepare.

Regards,
Bjorn