Re: [PATCH v8 1/5] staging: vc04_services: vchiq_arm: Add new bus type and device type

[Umang Jain]

Hi Greg,

(resending because it got bounced off a few lists due a line a copied 
with HTML tag..)

On 7/3/23 3:28 PM, Greg KH wrote:
> On Tue, Jun 27, 2023 at 10:16:24PM +0200, Umang Jain wrote:
> > The devices that the vchiq interface registers (bcm2835-audio,
> > bcm2835-camera) are implemented and exposed by the VC04 firmware.
> > The device tree describes the VC04 itself with the resources required
> > to communicate with it through a mailbox interface. However, the
> > vchiq interface registers these devices as platform devices. This
> > also means the specific drivers for these devices are getting
> > registered as platform drivers. This is not correct and a blatant
> > abuse of platform device/driver.
> >
> > Add a new bus type, vchiq_bus_type and device type (struct vchiq_device)
> > which will be used to migrate child devices that the vchiq interfaces
> > creates/registers from the platform device/driver.
> >
> > Signed-off-by: Umang Jain <umang.jain@xxxxxxxxxxxxxxxx>
> > ---
> >   drivers/staging/vc04_services/Makefile        |  1 +
> >   .../interface/vchiq_arm/vchiq_device.c        | 78 +++++++++++++++++++
> >   .../interface/vchiq_arm/vchiq_device.h        | 43 ++++++++++
> >   3 files changed, 122 insertions(+)
> >   create mode 100644 drivers/staging/vc04_services/interface/vchiq_arm/vchiq_device.c
> >   create mode 100644 drivers/staging/vc04_services/interface/vchiq_arm/vchiq_device.h
> >
> > diff --git a/drivers/staging/vc04_services/Makefile b/drivers/staging/vc04_services/Makefile
> > index 44794bdf6173..2d071e55e175 100644
> > --- a/drivers/staging/vc04_services/Makefile
> > +++ b/drivers/staging/vc04_services/Makefile
> > @@ -5,6 +5,7 @@ vchiq-objs := \
> >      interface/vchiq_arm/vchiq_core.o  \
> >      interface/vchiq_arm/vchiq_arm.o \
> >      interface/vchiq_arm/vchiq_debugfs.o \
> > +   interface/vchiq_arm/vchiq_device.o \
> >      interface/vchiq_arm/vchiq_connected.o \
> >   
> >   ifdef CONFIG_VCHIQ_CDEV
> > diff --git a/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_device.c b/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_device.c
> > new file mode 100644
> > index 000000000000..dff312e9735c
> > --- /dev/null
> > +++ b/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_device.c
> > @@ -0,0 +1,78 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +/*
> > + * vchiq_device.c - VCHIQ generic device and bus-type
> > + *
> > + * Copyright (c) 2023 Ideas On Board Oy
> > + */
> > +
> > +#include <linux/device/bus.h>
> > +#include <linux/slab.h>
> > +#include <linux/string.h>
> > +
> > +#include "vchiq_device.h"
> > +
> > +static int vchiq_bus_type_match(struct device *dev, struct device_driver *drv);
> > +
> > +struct bus_type vchiq_bus_type = {
> > +	.name   = "vchiq-bus",
> > +	.match  = vchiq_bus_type_match,
> > +};
> > +
> > +static int vchiq_bus_type_match(struct device *dev, struct device_driver *drv)
> > +{
> > +	if (dev->bus == &vchiq_bus_type &&
> > +	    strcmp(dev_name(dev), drv->name) == 0)
> > +		return 1;
> > +	return 0;
> > +}
> > +
> > +static void vchiq_device_release(struct device *dev)
> > +{
> > +	struct vchiq_device *device;
> > +
> > +	device = container_of(dev, struct vchiq_device, dev);
> > +	kfree(device);
> > +}
> > +
> > +int vchiq_device_register(struct device *parent, const char *name)
> > +{
> > +	struct vchiq_device *device = NULL;
> No need to set this to NULL.
>
> > +	int ret;
> > +
> > +	device = kzalloc(sizeof(*device), GFP_KERNEL);
> > +	if (!device)
> > +		return -ENOMEM;
> > +
> > +	device->dev.init_name = name;
> > +	device->dev.parent = parent;
> > +	device->dev.bus = &vchiq_bus_type;
> > +	device->dev.release = vchiq_device_release;
> > +
> > +	ret = device_register(&device->dev);
> > +	if (ret) {
> > +		put_device(&device->dev);
> > +		return -EINVAL;
> Why not return the error given to you?
>
> > +	}
> > +
> > +	return 0;
> You create a new device, shouldn't you return it?  How is it going to be
> looked up again?

So I haven't come across usage of the device other than unregistered 
right now. If you look at the platform_device static instances of 
bcm2835_camera, bcm2835-audio, they are also just used for unregistering 
the devices only.

So for unregistering devices on vchiq_bus - I am using the

    bus_for_each_dev(&vchiq_bus_type, NULL, NULL, vchiq_device_unregister);

in patch 3/5.

So basically I can return the instances right now, but there would be no 
user of those instances.
> > +}
> > +
> > +int vchiq_device_unregister(struct device *dev, void *data)
> You should be passing in a sruct vchiq_device *device here, right?
>
> And why the void pointer you do nothing with?

ack
> > +{
> > +	device_unregister(dev);
> > +	return 0;
> > +}
> No need to export this?

In the previous reviews you mentioned not to export it. The 
vchiq_bus_type is internal to the vchiq_driver to register it's child 
devices. I don't think any other module will be using this bus directly? 
Maybe I'm mistaken?
> thanks,
>
> greg k-h