On Thu 30 Nov 08:46 PST 2017, Loic Pallardy wrote:
> This patch parse existing carveout list to find a memory area
> matching on "vdev<vdev_id>buffer" name.
> If found, memory device will be used as parent for vdev creation, else
> rproc platform device will be used as today.
>
> Signed-off-by: Loic Pallardy <[email protected]>
> ---
> drivers/remoteproc/remoteproc_core.c | 13 +++++++++++++
> drivers/remoteproc/remoteproc_virtio.c | 2 +-
> include/linux/remoteproc.h | 1 +
> 3 files changed, 15 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/remoteproc/remoteproc_core.c b/drivers/remoteproc/remoteproc_core.c
> index 6b5e2b2..9c12319 100644
> --- a/drivers/remoteproc/remoteproc_core.c
> +++ b/drivers/remoteproc/remoteproc_core.c
> @@ -583,8 +583,11 @@ static int rproc_handle_vdev(struct rproc *rproc, struct fw_rsc_vdev *rsc,
> {
> struct device *dev = &rproc->dev;
> struct rproc_vdev *rvdev;
> + struct device *memdev = dev->parent;
> + struct rproc_mem_entry *carveout;
> int i, ret;
> static int index;
> + char name[16];
>
> /* make sure resource isn't truncated */
> if (sizeof(*rsc) + rsc->num_of_vrings * sizeof(struct fw_rsc_vdev_vring)
> @@ -637,6 +640,16 @@ static int rproc_handle_vdev(struct rproc *rproc, struct fw_rsc_vdev *rsc,
>
> list_add_tail(&rvdev->node, &rproc->rvdevs);
>
> + /* Find associated registered carveout. */
> + /* Try dedicated vdev buffer pool. */
> + snprintf(name, sizeof(name), "vdev%dbuffer", rvdev->index);
> + carveout = rproc_find_carveout_by_name(rproc, name);
> +
> + if (carveout && carveout->memdev)
> + memdev = &carveout->memdev->dev;
> +
> + rvdev->dev = memdev;
> +
> rproc_add_subdev(rproc, &rvdev->subdev,
> rproc_vdev_do_probe, rproc_vdev_do_remove);
>
> diff --git a/drivers/remoteproc/remoteproc_virtio.c b/drivers/remoteproc/remoteproc_virtio.c
> index 2946348..1f7a444 100644
> --- a/drivers/remoteproc/remoteproc_virtio.c
> +++ b/drivers/remoteproc/remoteproc_virtio.c
> @@ -303,7 +303,7 @@ static void rproc_virtio_dev_release(struct device *dev)
> int rproc_add_virtio_dev(struct rproc_vdev *rvdev, int id)
> {
> struct rproc *rproc = rvdev->rproc;
> - struct device *dev = &rproc->dev;
> + struct device *dev = rvdev->dev;
This will cause the device structure to change shape based on there
being a match of a carveout or not.
I also think it's preferable to limit the life cycle of the allocations
in this memory region to a single start->stop cycle, rather than
boot->shutdown.
So I think it makes more sense to use the vdev->dev and
dmam_declare_coherent_memory on this. But as in the previous patch this
can't be a carveout that has been remapped already.
A somewhat unrelated topic to this is the ability to associate DT nodes
to rpmsg devices (I do this for the Qualcomm children), in this case we
would have a DT node per vdev under the remoteproc, perhaps it would
make more sense to introduce that and put the memory-region in that
node. Only thin that comes to mind is that we still need to match a
carveout in the resource table, in order to communicate the buffer
region to the remote side for your memory protection purposes.
Regards,
Bjorn
> -----Original Message-----
> From: Bjorn Andersson [mailto:[email protected]]
> Sent: Thursday, December 14, 2017 6:33 AM
> To: Loic PALLARDY <[email protected]>
> Cc: [email protected]; [email protected]; linux-
> [email protected]; Arnaud POULIQUEN <[email protected]>;
> [email protected]
> Subject: Re: [PATCH v2 13/16] remoteproc: look-up memory-device for virtio
> device allocation
>
> On Thu 30 Nov 08:46 PST 2017, Loic Pallardy wrote:
>
> > This patch parse existing carveout list to find a memory area
> > matching on "vdev<vdev_id>buffer" name.
> > If found, memory device will be used as parent for vdev creation, else
> > rproc platform device will be used as today.
> >
> > Signed-off-by: Loic Pallardy <[email protected]>
> > ---
> > drivers/remoteproc/remoteproc_core.c | 13 +++++++++++++
> > drivers/remoteproc/remoteproc_virtio.c | 2 +-
> > include/linux/remoteproc.h | 1 +
> > 3 files changed, 15 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/remoteproc/remoteproc_core.c
> b/drivers/remoteproc/remoteproc_core.c
> > index 6b5e2b2..9c12319 100644
> > --- a/drivers/remoteproc/remoteproc_core.c
> > +++ b/drivers/remoteproc/remoteproc_core.c
> > @@ -583,8 +583,11 @@ static int rproc_handle_vdev(struct rproc *rproc,
> struct fw_rsc_vdev *rsc,
> > {
> > struct device *dev = &rproc->dev;
> > struct rproc_vdev *rvdev;
> > + struct device *memdev = dev->parent;
> > + struct rproc_mem_entry *carveout;
> > int i, ret;
> > static int index;
> > + char name[16];
> >
> > /* make sure resource isn't truncated */
> > if (sizeof(*rsc) + rsc->num_of_vrings * sizeof(struct
> fw_rsc_vdev_vring)
> > @@ -637,6 +640,16 @@ static int rproc_handle_vdev(struct rproc *rproc,
> struct fw_rsc_vdev *rsc,
> >
> > list_add_tail(&rvdev->node, &rproc->rvdevs);
> >
> > + /* Find associated registered carveout. */
> > + /* Try dedicated vdev buffer pool. */
> > + snprintf(name, sizeof(name), "vdev%dbuffer", rvdev->index);
> > + carveout = rproc_find_carveout_by_name(rproc, name);
> > +
> > + if (carveout && carveout->memdev)
> > + memdev = &carveout->memdev->dev;
> > +
> > + rvdev->dev = memdev;
> > +
> > rproc_add_subdev(rproc, &rvdev->subdev,
> > rproc_vdev_do_probe, rproc_vdev_do_remove);
> >
> > diff --git a/drivers/remoteproc/remoteproc_virtio.c
> b/drivers/remoteproc/remoteproc_virtio.c
> > index 2946348..1f7a444 100644
> > --- a/drivers/remoteproc/remoteproc_virtio.c
> > +++ b/drivers/remoteproc/remoteproc_virtio.c
> > @@ -303,7 +303,7 @@ static void rproc_virtio_dev_release(struct device
> *dev)
> > int rproc_add_virtio_dev(struct rproc_vdev *rvdev, int id)
> > {
> > struct rproc *rproc = rvdev->rproc;
> > - struct device *dev = &rproc->dev;
> > + struct device *dev = rvdev->dev;
>
> This will cause the device structure to change shape based on there
> being a match of a carveout or not.
>
>
> I also think it's preferable to limit the life cycle of the allocations
> in this memory region to a single start->stop cycle, rather than
> boot->shutdown.
>
> So I think it makes more sense to use the vdev->dev and
> dmam_declare_coherent_memory on this. But as in the previous patch this
> can't be a carveout that has been remapped already.
>
Yes it is the main issue. Rproc_handle_carveout function will process it before, except if we create a new status to not allocate a carveout and provide function to update carevout in resource table...
>
> A somewhat unrelated topic to this is the ability to associate DT nodes
> to rpmsg devices (I do this for the Qualcomm children), in this case we
> would have a DT node per vdev under the remoteproc, perhaps it would
> make more sense to introduce that and put the memory-region in that
> node. Only thin that comes to mind is that we still need to match a
> carveout in the resource table, in order to communicate the buffer
> region to the remote side for your memory protection purposes.
>
OK I'll have a look to Qualcom DT nodes. But as it is SW, is DT the right place?
But it will help to decide if there is a dedicated memory region or not for virtio device.
I'll investigate...
Regards,
Loic
> Regards,
> Bjorn