Skip to contentSkip to conversation

s390: vfio_ap: link the vfio_ap devices to the vfio_ap bus subsystem

PATCH Superseded Pierre Morel 6 participants 20 messages Feb 18, 2019 – Feb 21, 2019

Subsystems (3)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

v2 cover s390: vfio_ap: link the vfio_ap devices to the vfio_ap bus subsystem

This is the second version of this patch.
The first version was part of a patch series on VFIO AP IRQ.

The goal of this patch is to standardize the device-driver interface for the VFIO_AP ap_matrix_device to satisfy user-land tools working on hot-plug (UDEV/LIBVIRT).

Christian Borntraeger reported libvirt looping when a matrix device was available before the libvirt start.

Marc Hartmayer debugged this and circumvented this in libvirt:
https://www.redhat.com/archives/libvir-list/2019-February/msg00837.html

Pierre Morel (1):
s390: vfio_ap: link the vfio_ap devices to the vfio_ap bus subsystem

 drivers/s390/crypto/vfio_ap_drv.c     | 48 +++++++++++++++++++++++++++++------
 drivers/s390/crypto/vfio_ap_ops.c     |  4 +--
 drivers/s390/crypto/vfio_ap_private.h |  1 +
 3 files changed, 43 insertions(+), 10 deletions(-)
Signature · 1 line
2.7.4
#1 lore raw

PATCH v2 1/1 s390: vfio_ap: link the vfio_ap devices to the vfio_ap bus subsystem

Libudev relies on having a subsystem link for non-root devices. To avoid libudev (and potentially other userspace tools) choking on the matrix device let us introduce a vfio_ap bus and with that the vfio_ap bus subsytem, and make the matrix device reside within it.

Doing this we need to suppress the forced link from the matrix device to the vfio_ap driver and we suppress the device_type we do not need anymore.

Since the associated matrix driver is not the vfio_ap driver any more, we have to change the search for the devices on the vfio_ap driver in the function vfio_ap_verify_queue_reserved.

---
 drivers/s390/crypto/vfio_ap_drv.c     | 48 +++++++++++++++++++++++++++++------
 drivers/s390/crypto/vfio_ap_ops.c     |  4 +--
 drivers/s390/crypto/vfio_ap_private.h |  1 +
 3 files changed, 43 insertions(+), 10 deletions(-)
drivers/s390/crypto/vfio_ap_drv.c +40 -8
index 31c6c84..8e45559 100644
@@ -24,10 +24,6 @@ MODULE_LICENSE("GPL v2");
 
 static struct ap_driver vfio_ap_drv;
 
-static struct device_type vfio_ap_dev_type = {
-	.name = VFIO_AP_DEV_TYPE_NAME,
-};
-
 struct ap_matrix_dev *matrix_dev;
 
 /* Only type 10 adapters (CEX4 and later) are supported
@@ -62,6 +58,27 @@ static void vfio_ap_matrix_dev_release(struct device *dev)
 	kfree(matrix_dev);
 }
 
+static int matrix_bus_match(struct device *dev, struct device_driver *drv)
+{
+	return 1;
+}
+
+static struct bus_type matrix_bus = {
+	.name = "vfio_ap",
+	.match = &matrix_bus_match,
+};
+
+static int matrix_probe(struct device *dev)
+{
+	return 0;
+}
+
+static struct device_driver matrix_driver = {
+	.name = "vfio_ap",
+	.bus = &matrix_bus,
+	.probe = matrix_probe,
+};
+
 static int vfio_ap_matrix_dev_create(void)
 {
 	int ret;
@@ -71,6 +88,10 @@ static int vfio_ap_matrix_dev_create(void)
 	if (IS_ERR(root_device))
 		return PTR_ERR(root_device);
 
+	ret = bus_register(&matrix_bus);
+	if (ret)
+		goto bus_register_err;
+
 	matrix_dev = kzalloc(sizeof(*matrix_dev), GFP_KERNEL);
 	if (!matrix_dev) {
 		ret = -ENOMEM;
@@ -87,30 +108,41 @@ static int vfio_ap_matrix_dev_create(void)
 	mutex_init(&matrix_dev->lock);
 	INIT_LIST_HEAD(&matrix_dev->mdev_list);
 
-	matrix_dev->device.type = &vfio_ap_dev_type;
 	dev_set_name(&matrix_dev->device, "%s", VFIO_AP_DEV_NAME);
 	matrix_dev->device.parent = root_device;
+	matrix_dev->device.bus = &matrix_bus;
 	matrix_dev->device.release = vfio_ap_matrix_dev_release;
-	matrix_dev->device.driver = &vfio_ap_drv.driver;
+	matrix_dev->vfio_ap_drv = &vfio_ap_drv;
 
 	ret = device_register(&matrix_dev->device);
 	if (ret)
 		goto matrix_reg_err;
 
+	ret = driver_register(&matrix_driver);
+	if (ret)
+		goto matrix_drv_err;
+
 	return 0;
 
+matrix_drv_err:
+	device_unregister(&matrix_dev->device);
 matrix_reg_err:
 	put_device(&matrix_dev->device);
 matrix_alloc_err:
+	bus_unregister(&matrix_bus);
+bus_register_err:
 	root_device_unregister(root_device);
-
 	return ret;
 }
 
 static void vfio_ap_matrix_dev_destroy(void)
 {
+	struct device *root_device = matrix_dev->device.parent;
+
+	driver_unregister(&matrix_driver);
 	device_unregister(&matrix_dev->device);
-	root_device_unregister(matrix_dev->device.parent);
+	bus_unregister(&matrix_bus);
+	root_device_unregister(root_device);
 }
 
 static int __init vfio_ap_init(void)
drivers/s390/crypto/vfio_ap_ops.c +2 -2
index 272ef42..900b9cf 100644
@@ -198,8 +198,8 @@ static int vfio_ap_verify_queue_reserved(unsigned long *apid,
 	qres.apqi = apqi;
 	qres.reserved = false;
 
-	ret = driver_for_each_device(matrix_dev->device.driver, NULL, &qres,
-				     vfio_ap_has_queue);
+	ret = driver_for_each_device(&matrix_dev->vfio_ap_drv->driver, NULL,
+				     &qres, vfio_ap_has_queue);
 	if (ret)
 		return ret;
 
drivers/s390/crypto/vfio_ap_private.h +1 -0
index 5675492..76b7f98 100644
@@ -40,6 +40,7 @@ struct ap_matrix_dev {
 	struct ap_config_info info;
 	struct list_head mdev_list;
 	struct mutex lock;
+	struct ap_driver  *vfio_ap_drv;
 };
 
 extern struct ap_matrix_dev *matrix_dev;
Signature · 1 line
2.7.4
Tested-by on v2 1/1
#2 lore raw
Quoted 9 lines · Pierre Morel

Libudev relies on having a subsystem link for non-root devices. To avoid libudev (and potentially other userspace tools) choking on the matrix device let us introduce a vfio_ap bus and with that the vfio_ap bus subsytem, and make the matrix device reside within it.

Doing this we need to suppress the forced link from the matrix device to the vfio_ap driver and we suppress the device_type we do not need anymore.

Quoted 6 lines

Since the associated matrix driver is not the vfio_ap driver any more, we have to change the search for the devices on the vfio_ap driver in the function vfio_ap_verify_queue_reserved.

Can you also add a "Fixes:" tag

Would be good to have someone with device experience reviewing the changed.

Quoted 135 lines
---
 drivers/s390/crypto/vfio_ap_drv.c     | 48 +++++++++++++++++++++++++++++------
 drivers/s390/crypto/vfio_ap_ops.c     |  4 +--
 drivers/s390/crypto/vfio_ap_private.h |  1 +
 3 files changed, 43 insertions(+), 10 deletions(-)
drivers/s390/crypto/vfio_ap_drv.c +40 -8
index 31c6c84..8e45559 100644
@@ -24,10 +24,6 @@ MODULE_LICENSE("GPL v2");
 
 static struct ap_driver vfio_ap_drv;
 
-static struct device_type vfio_ap_dev_type = {
-	.name = VFIO_AP_DEV_TYPE_NAME,
-};
-
 struct ap_matrix_dev *matrix_dev;
 
 /* Only type 10 adapters (CEX4 and later) are supported
@@ -62,6 +58,27 @@ static void vfio_ap_matrix_dev_release(struct device *dev)
 	kfree(matrix_dev);
 }
 
+static int matrix_bus_match(struct device *dev, struct device_driver *drv)
+{
+	return 1;
+}
+
+static struct bus_type matrix_bus = {
+	.name = "vfio_ap",
+	.match = &matrix_bus_match,
+};
+
+static int matrix_probe(struct device *dev)
+{
+	return 0;
+}
+
+static struct device_driver matrix_driver = {
+	.name = "vfio_ap",
+	.bus = &matrix_bus,
+	.probe = matrix_probe,
+};
+
 static int vfio_ap_matrix_dev_create(void)
 {
 	int ret;
@@ -71,6 +88,10 @@ static int vfio_ap_matrix_dev_create(void)
 	if (IS_ERR(root_device))
 		return PTR_ERR(root_device);
 
+	ret = bus_register(&matrix_bus);
+	if (ret)
+		goto bus_register_err;
+
 	matrix_dev = kzalloc(sizeof(*matrix_dev), GFP_KERNEL);
 	if (!matrix_dev) {
 		ret = -ENOMEM;
@@ -87,30 +108,41 @@ static int vfio_ap_matrix_dev_create(void)
 	mutex_init(&matrix_dev->lock);
 	INIT_LIST_HEAD(&matrix_dev->mdev_list);
 
-	matrix_dev->device.type = &vfio_ap_dev_type;
 	dev_set_name(&matrix_dev->device, "%s", VFIO_AP_DEV_NAME);
 	matrix_dev->device.parent = root_device;
+	matrix_dev->device.bus = &matrix_bus;
 	matrix_dev->device.release = vfio_ap_matrix_dev_release;
-	matrix_dev->device.driver = &vfio_ap_drv.driver;
+	matrix_dev->vfio_ap_drv = &vfio_ap_drv;
 
 	ret = device_register(&matrix_dev->device);
 	if (ret)
 		goto matrix_reg_err;
 
+	ret = driver_register(&matrix_driver);
+	if (ret)
+		goto matrix_drv_err;
+
 	return 0;
 
+matrix_drv_err:
+	device_unregister(&matrix_dev->device);
 matrix_reg_err:
 	put_device(&matrix_dev->device);
 matrix_alloc_err:
+	bus_unregister(&matrix_bus);
+bus_register_err:
 	root_device_unregister(root_device);
-
 	return ret;
 }
 
 static void vfio_ap_matrix_dev_destroy(void)
 {
+	struct device *root_device = matrix_dev->device.parent;
+
+	driver_unregister(&matrix_driver);
 	device_unregister(&matrix_dev->device);
-	root_device_unregister(matrix_dev->device.parent);
+	bus_unregister(&matrix_bus);
+	root_device_unregister(root_device);
 }
 
 static int __init vfio_ap_init(void)
drivers/s390/crypto/vfio_ap_ops.c +2 -2
index 272ef42..900b9cf 100644
@@ -198,8 +198,8 @@ static int vfio_ap_verify_queue_reserved(unsigned long *apid,
 	qres.apqi = apqi;
 	qres.reserved = false;
 
-	ret = driver_for_each_device(matrix_dev->device.driver, NULL, &qres,
-				     vfio_ap_has_queue);
+	ret = driver_for_each_device(&matrix_dev->vfio_ap_drv->driver, NULL,
+				     &qres, vfio_ap_has_queue);
 	if (ret)
 		return ret;
 
drivers/s390/crypto/vfio_ap_private.h +1 -0
index 5675492..76b7f98 100644
@@ -40,6 +40,7 @@ struct ap_matrix_dev {
 	struct ap_config_info info;
 	struct list_head mdev_list;
 	struct mutex lock;
+	struct ap_driver  *vfio_ap_drv;
 };
 
 extern struct ap_matrix_dev *matrix_dev;
Reviewed-by on v2 1/1
#3 lore raw
In reply to Pierre Morel
Quoted 7 lines · On Mon, 18 Feb 2019 19:08:48 +0100 Pierre Morel <[email protected]> wrote:

Libudev relies on having a subsystem link for non-root devices. To avoid libudev (and potentially other userspace tools) choking on the matrix device let us introduce a vfio_ap bus and with that the vfio_ap bus subsytem, and make the matrix device reside within it.

Doing this we need to suppress the forced link from the matrix device to the vfio_ap driver and we suppress the device_type we do not need

s/suppress/remove/ ?

Quoted 14 lines

anymore.

Since the associated matrix driver is not the vfio_ap driver any more, we have to change the search for the devices on the vfio_ap driver in the function vfio_ap_verify_queue_reserved.

---
 drivers/s390/crypto/vfio_ap_drv.c     | 48 +++++++++++++++++++++++++++++------
 drivers/s390/crypto/vfio_ap_ops.c     |  4 +--
 drivers/s390/crypto/vfio_ap_private.h |  1 +
 3 files changed, 43 insertions(+), 10 deletions(-)

(...)

Quoted 8 lines
drivers/s390/crypto/vfio_ap_drv.c +4 -0
@@ -62,6 +58,27 @@ static void vfio_ap_matrix_dev_release(struct device *dev)
 	kfree(matrix_dev);
 }
 
+static int matrix_bus_match(struct device *dev, struct device_driver *drv)
+{
+	return 1;
+}

drivers/s390/crypto/vfio_ap_drv.c static void vfio_ap_matrix_dev_release(struct device *dev) · line 74

Quoted 10 lines
drivers/s390/crypto/vfio_ap_drv.c +10 -0
+
+static struct bus_type matrix_bus = {
+	.name = "vfio_ap",
+	.match = &matrix_bus_match,
+};
+
+static int matrix_probe(struct device *dev)
+{
+	return 0;
+}

I don't think you need this (the important function is the match function of the bus).

drivers/s390/crypto/vfio_ap_drv.c static void vfio_ap_matrix_dev_release(struct device *dev) · line 84

Quoted 10 lines
drivers/s390/crypto/vfio_ap_drv.c +7 -0
+
+static struct device_driver matrix_driver = {
+	.name = "vfio_ap",
+	.bus = &matrix_bus,
+	.probe = matrix_probe,
+};
+
 static int vfio_ap_matrix_dev_create(void)
 {
 	int ret;

It's a bit annoying that we need to introduce a bus that basically does nothing, but I think this looks sane.

#4 lore raw
Quoted 43 lines · Cornelia Huck
Quoted 7 lines · On Mon, 18 Feb 2019 19:08:48 +0100 Pierre Morel <[email protected]> wrote:

Libudev relies on having a subsystem link for non-root devices. To avoid libudev (and potentially other userspace tools) choking on the matrix device let us introduce a vfio_ap bus and with that the vfio_ap bus subsytem, and make the matrix device reside within it.

Doing this we need to suppress the forced link from the matrix device to the vfio_ap driver and we suppress the device_type we do not need

s/suppress/remove/ ?

Quoted 14 lines

anymore.

Since the associated matrix driver is not the vfio_ap driver any more, we have to change the search for the devices on the vfio_ap driver in the function vfio_ap_verify_queue_reserved.

---
 drivers/s390/crypto/vfio_ap_drv.c     | 48 +++++++++++++++++++++++++++++------
 drivers/s390/crypto/vfio_ap_ops.c     |  4 +--
 drivers/s390/crypto/vfio_ap_private.h |  1 +
 3 files changed, 43 insertions(+), 10 deletions(-)

(...)

Quoted 13 lines
(quoted diff) +9 -0
@@ -62,6 +58,27 @@ static void vfio_ap_matrix_dev_release(struct device *dev)
 	kfree(matrix_dev);
 }
 
+static int matrix_bus_match(struct device *dev, struct device_driver *drv)
+{
+	return 1;
+}
+
+static struct bus_type matrix_bus = {
+	.name = "vfio_ap",
+	.match = &matrix_bus_match,
+};
Quoted 8 lines

Review static void vfio_ap_matrix_dev_release(struct device *dev) · line 74

Quoted 5 lines
(quoted diff) +5 -0
+
+static int matrix_probe(struct device *dev)
+{
+	return 0;
+}

I don't think you need this (the important function is the match function of the bus).

Yes, with

drivers/s390/crypto/vfio_ap_drv.c +0 -6
index 8e45559795429..8ceec41afe322 100644
@@ -68,15 +68,9 @@ static struct bus_type matrix_bus = {
        .match = &matrix_bus_match,
 };
 
-static int matrix_probe(struct device *dev)
-{
-       return 0;
-}
-
 static struct device_driver matrix_driver = {
        .name = "vfio_ap",
        .bus = &matrix_bus,
-       .probe = matrix_probe,
 };
 
 static int vfio_ap_matrix_dev_create(void)

on top things still look fine.

Quoted 17 lines

Review static void vfio_ap_matrix_dev_release(struct device *dev) · line 84

Quoted 11 lines
(quoted diff) +7 -0
+
+static struct device_driver matrix_driver = {
+	.name = "vfio_ap",
+	.bus = &matrix_bus,
+	.probe = matrix_probe,
+};
+
 static int vfio_ap_matrix_dev_create(void)
 {
 	int ret;

It's a bit annoying that we need to introduce a bus that basically does nothing, but I think this looks sane.

#5 lore raw
Quoted 6 lines · Christian Borntraeger
Quoted 3 lines · Cornelia Huck

On Mon, 18 Feb 2019 19:08:48 +0100
Pierre Morel <[email protected]> wrote:

...snip...

Quoted 36 lines
Quoted 4 lines
Quoted 4 lines
(quoted diff) +4 -0
+static struct bus_type matrix_bus = {
+	.name = "vfio_ap",
+	.match = &matrix_bus_match,
+};
Quoted 8 lines

Review

Quoted 5 lines
(quoted diff) +5 -0
+
+static int matrix_probe(struct device *dev)
+{
+	return 0;
+}

I don't think you need this (the important function is the match function of the bus).

Yes, with

drivers/s390/crypto/vfio_ap_drv.c +0 -6
index 8e45559795429..8ceec41afe322 100644
@@ -68,15 +68,9 @@ static struct bus_type matrix_bus = {
         .match = &matrix_bus_match,
  };
 
-static int matrix_probe(struct device *dev)
-{
-       return 0;
-}
-
  static struct device_driver matrix_driver = {
         .name = "vfio_ap",
         .bus = &matrix_bus,
-       .probe = matrix_probe,
  };
 
  static int vfio_ap_matrix_dev_create(void)
Quoted 1 line

on top things still look fine.

OK, simpler is better, I remove the probe.

Thanks,
Pierre

Signature · 2 lines
Pierre Morel
Linux/KVM/QEMU in Böblingen - Germany
maintainer Acked-by on v2 1/1
#6 lore raw
In reply to Pierre Morel
Quoted 27 lines · On Mon, 18 Feb 2019 19:08:48 +0100 Pierre Morel <[email protected]> wrote:

Libudev relies on having a subsystem link for non-root devices. To avoid libudev (and potentially other userspace tools) choking on the matrix device let us introduce a vfio_ap bus and with that the vfio_ap bus subsytem, and make the matrix device reside within it.

Doing this we need to suppress the forced link from the matrix device to the vfio_ap driver and we suppress the device_type we do not need anymore.

Since the associated matrix driver is not the vfio_ap driver any more, we have to change the search for the devices on the vfio_ap driver in the function vfio_ap_verify_queue_reserved.

---
 drivers/s390/crypto/vfio_ap_drv.c     | 48 +++++++++++++++++++++++++++++------
 drivers/s390/crypto/vfio_ap_ops.c     |  4 +--
 drivers/s390/crypto/vfio_ap_private.h |  1 +
 3 files changed, 43 insertions(+), 10 deletions(-)
drivers/s390/crypto/vfio_ap_drv.c +0 -0
index 31c6c84..8e45559 100644

drivers/s390/crypto/vfio_ap_drv.c MODULE_LICENSE("GPL v2"); · line 30

Quoted 7 lines
drivers/s390/crypto/vfio_ap_drv.c +0 -4
@@ -24,10 +24,6 @@ MODULE_LICENSE("GPL v2");
 
 static struct ap_driver vfio_ap_drv;
 
-static struct device_type vfio_ap_dev_type = {
-	.name = VFIO_AP_DEV_TYPE_NAME,
-};
-

Would have been nice had you mentioned this in the commit message -- vfio_ap_dev_type was useless. And you should have removed the define of VFIO_AP_DEV_TYPE_NAME a couple of lines above this.

Quoted 18 lines
drivers/s390/crypto/vfio_ap_drv.c +11 -0
 struct ap_matrix_dev *matrix_dev;
 
 /* Only type 10 adapters (CEX4 and later) are supported
@@ -62,6 +58,27 @@ static void vfio_ap_matrix_dev_release(struct device *dev)
 	kfree(matrix_dev);
 }
 
+static int matrix_bus_match(struct device *dev, struct device_driver *drv)
+{
+	return 1;
+}
+
+static struct bus_type matrix_bus = {
+	.name = "vfio_ap",
+	.match = &matrix_bus_match,
+};
+
+static int matrix_probe(struct device *dev)

drivers/s390/crypto/vfio_ap_drv.c static void vfio_ap_matrix_dev_release(struct device *dev) · line 81

Quoted 10 lines
drivers/s390/crypto/vfio_ap_drv.c +10 -0
+{
+	return 0;
+}
+
+static struct device_driver matrix_driver = {
+	.name = "vfio_ap",
+	.bus = &matrix_bus,
+	.probe = matrix_probe,
+};
+

With the changes suggested by Christian:

[..]

maintainer
#7 lore raw
In reply to Pierre Morel
Quoted 38 lines · Pierre Morel

Libudev relies on having a subsystem link for non-root devices. To avoid libudev (and potentially other userspace tools) choking on the matrix device let us introduce a vfio_ap bus and with that the vfio_ap bus subsytem, and make the matrix device reside within it.

Doing this we need to suppress the forced link from the matrix device to the vfio_ap driver and we suppress the device_type we do not need anymore.

Since the associated matrix driver is not the vfio_ap driver any more, we have to change the search for the devices on the vfio_ap driver in the function vfio_ap_verify_queue_reserved.

---
  drivers/s390/crypto/vfio_ap_drv.c     | 48 +++++++++++++++++++++++++++++------
  drivers/s390/crypto/vfio_ap_ops.c     |  4 +--
  drivers/s390/crypto/vfio_ap_private.h |  1 +
  3 files changed, 43 insertions(+), 10 deletions(-)
drivers/s390/crypto/vfio_ap_drv.c +0 -4
index 31c6c84..8e45559 100644
@@ -24,10 +24,6 @@ MODULE_LICENSE("GPL v2");
 
  static struct ap_driver vfio_ap_drv;
 
-static struct device_type vfio_ap_dev_type = {
-	.name = VFIO_AP_DEV_TYPE_NAME,
-};
-
  struct ap_matrix_dev *matrix_dev;
 
  /* Only type 10 adapters (CEX4 and later) are supported

drivers/s390/crypto/vfio_ap_drv.c static void vfio_ap_matrix_dev_release(struct device *dev) · line 63

Quoted 6 lines
drivers/s390/crypto/vfio_ap_drv.c +3 -0
@@ -62,6 +58,27 @@ static void vfio_ap_matrix_dev_release(struct device *dev)
  	kfree(matrix_dev);
  }
 
+static int matrix_bus_match(struct device *dev, struct device_driver *drv)
+{
+	return 1;

I think we should verify the following:

* dev == matrix_dev->device
* drv == &matrix_driver

The model employed is for the matrix device to be a singleton, so I think we should verify that the matrix device and driver defined herein ought to be the only possible choices for a match. Of course, doing so will require some restructuring of this patch.

Quoted 4 lines
drivers/s390/crypto/vfio_ap_drv.c +4 -0
+}
+
+static struct bus_type matrix_bus = {
+	.name = "vfio_ap",

drivers/s390/crypto/vfio_ap_drv.c static void vfio_ap_matrix_dev_release(struct device *dev) · line 77

Quoted 10 lines
drivers/s390/crypto/vfio_ap_drv.c +10 -0
+	.match = &matrix_bus_match,
+};
+
+static int matrix_probe(struct device *dev)
+{
+	return 0;
+}
+
+static struct device_driver matrix_driver = {
+	.name = "vfio_ap",

This is the same name used for the original device driver. I think this driver ought to have a different name to avoid confusion. How about vfio_ap_matrix or some other name to differentiate the two.

drivers/s390/crypto/vfio_ap_drv.c static void vfio_ap_matrix_dev_release(struct device *dev) · line 79

Quoted 2 lines
drivers/s390/crypto/vfio_ap_drv.c +2 -0
+	.bus = &matrix_bus,
+	.probe = matrix_probe,

I would add:

	.suppress_bind_attrs = true;

This will remove the sysfs bind/unbind interfaces. Since there is only one matrix device and it's lifecycle is controlled herein, there is no sense in allowing a root user to bind/unbind it.

Quoted 90 lines
drivers/s390/crypto/vfio_ap_drv.c +21 -4
+};
+
  static int vfio_ap_matrix_dev_create(void)
  {
  	int ret;
@@ -71,6 +88,10 @@ static int vfio_ap_matrix_dev_create(void)
  	if (IS_ERR(root_device))
  		return PTR_ERR(root_device);
 
+	ret = bus_register(&matrix_bus);
+	if (ret)
+		goto bus_register_err;
+
  	matrix_dev = kzalloc(sizeof(*matrix_dev), GFP_KERNEL);
  	if (!matrix_dev) {
  		ret = -ENOMEM;
@@ -87,30 +108,41 @@ static int vfio_ap_matrix_dev_create(void)
  	mutex_init(&matrix_dev->lock);
  	INIT_LIST_HEAD(&matrix_dev->mdev_list);
 
-	matrix_dev->device.type = &vfio_ap_dev_type;
  	dev_set_name(&matrix_dev->device, "%s", VFIO_AP_DEV_NAME);
  	matrix_dev->device.parent = root_device;
+	matrix_dev->device.bus = &matrix_bus;
  	matrix_dev->device.release = vfio_ap_matrix_dev_release;
-	matrix_dev->device.driver = &vfio_ap_drv.driver;
+	matrix_dev->vfio_ap_drv = &vfio_ap_drv;
 
  	ret = device_register(&matrix_dev->device);
  	if (ret)
  		goto matrix_reg_err;
 
+	ret = driver_register(&matrix_driver);
+	if (ret)
+		goto matrix_drv_err;
+
  	return 0;
 
+matrix_drv_err:
+	device_unregister(&matrix_dev->device);
  matrix_reg_err:
  	put_device(&matrix_dev->device);
  matrix_alloc_err:
+	bus_unregister(&matrix_bus);
+bus_register_err:
  	root_device_unregister(root_device);
-
  	return ret;
  }
 
  static void vfio_ap_matrix_dev_destroy(void)
  {
+	struct device *root_device = matrix_dev->device.parent;
+
+	driver_unregister(&matrix_driver);
  	device_unregister(&matrix_dev->device);
-	root_device_unregister(matrix_dev->device.parent);
+	bus_unregister(&matrix_bus);
+	root_device_unregister(root_device);
  }
 
  static int __init vfio_ap_init(void)
drivers/s390/crypto/vfio_ap_ops.c +2 -2
index 272ef42..900b9cf 100644
@@ -198,8 +198,8 @@ static int vfio_ap_verify_queue_reserved(unsigned long *apid,
  	qres.apqi = apqi;
  	qres.reserved = false;
 
-	ret = driver_for_each_device(matrix_dev->device.driver, NULL, &qres,
-				     vfio_ap_has_queue);
+	ret = driver_for_each_device(&matrix_dev->vfio_ap_drv->driver, NULL,
+				     &qres, vfio_ap_has_queue);
  	if (ret)
  		return ret;
 
drivers/s390/crypto/vfio_ap_private.h +1 -0
index 5675492..76b7f98 100644
@@ -40,6 +40,7 @@ struct ap_matrix_dev {
  	struct ap_config_info info;
  	struct list_head mdev_list;
  	struct mutex lock;
+	struct ap_driver  *vfio_ap_drv;
  };
 
  extern struct ap_matrix_dev *matrix_dev;
#8 lore raw
Quoted 47 lines · Tony Krowiak
Quoted 36 lines · Pierre Morel

Libudev relies on having a subsystem link for non-root devices. To avoid libudev (and potentially other userspace tools) choking on the matrix device let us introduce a vfio_ap bus and with that the vfio_ap bus subsytem, and make the matrix device reside within it.

Doing this we need to suppress the forced link from the matrix device to the vfio_ap driver and we suppress the device_type we do not need anymore.

Since the associated matrix driver is not the vfio_ap driver any more, we have to change the search for the devices on the vfio_ap driver in the function vfio_ap_verify_queue_reserved.

---
  drivers/s390/crypto/vfio_ap_drv.c     | 48
+++++++++++++++++++++++++++++------
  drivers/s390/crypto/vfio_ap_ops.c     |  4 +--
  drivers/s390/crypto/vfio_ap_private.h |  1 +

  3 files changed, 43 insertions(+), 10 deletions(-)
drivers/s390/crypto/vfio_ap_drv.c +0 -4
index 31c6c84..8e45559 100644
@@ -24,10 +24,6 @@ MODULE_LICENSE("GPL v2");
   static struct ap_driver vfio_ap_drv;
-static struct device_type vfio_ap_dev_type = {
-    .name = VFIO_AP_DEV_TYPE_NAME,
-};
-
   struct ap_matrix_dev *matrix_dev;
   /* Only type 10 adapters (CEX4 and later) are supported

drivers/s390/crypto/vfio_ap_drv.c static void vfio_ap_matrix_dev_release(struct · line 64

Quoted 7 lines
drivers/s390/crypto/vfio_ap_drv.c +3 -0
@@ -62,6 +58,27 @@ static void vfio_ap_matrix_dev_release(struct 
 device *dev)
       kfree(matrix_dev);
   }
+static int matrix_bus_match(struct device *dev, struct device_driver 
 *drv)
+{
+    return 1;

I think we should verify the following:

Quoted 7 lines

* dev == matrix_dev->device
* drv == &matrix_driver

The model employed is for the matrix device to be a singleton, so I think we should verify that the matrix device and driver defined herein ought to be the only possible choices for a match. Of course, doing so will require some restructuring of this patch.

I think Conny already answered this question.

Quoted 12 lines
Quoted 12 lines
drivers/s390/crypto/vfio_ap_drv.c +11 -0
+}
+
+static struct bus_type matrix_bus = {
+    .name = "vfio_ap",
+    .match = &matrix_bus_match,
+};
+
+static int matrix_probe(struct device *dev)
+{
+    return 0;
+}
Quoted 8 lines

drivers/s390/crypto/vfio_ap_drv.c static void vfio_ap_matrix_dev_release(struct device *dev) · line 77

Quoted 3 lines
drivers/s390/crypto/vfio_ap_drv.c +3 -0
+
+static struct device_driver matrix_driver = {
+    .name = "vfio_ap",

This is the same name used for the original device driver. I think this driver ought to have a different name to avoid confusion. How about vfio_ap_matrix or some other name to differentiate the two.

I would like too, but changing this will change the path to the mediated device supported type.

Quoted 11 lines

drivers/s390/crypto/vfio_ap_drv.c static void vfio_ap_matrix_dev_release(struct device *dev) · line 79

Quoted 3 lines
drivers/s390/crypto/vfio_ap_drv.c +2 -0
+    .bus = &matrix_bus,
+    .probe = matrix_probe,

I would add:
    .suppress_bind_attrs = true;

This will remove the sysfs bind/unbind interfaces. Since there is only one matrix device and it's lifecycle is controlled herein, there is no sense in allowing a root user to bind/unbind it.

OTOH bind/unbind has no impact.
If no one else ask for this I will not change what has already been reviewed by Conny and Christian.

Regards,
Pierre

Signature · 2 lines
Pierre Morel
Linux/KVM/QEMU in Böblingen - Germany
#9 lore raw
Quoted 52 lines · On Tue, 19 Feb 2019 22:31:17 +0100 Pierre Morel <[email protected]> wrote:
Quoted 51 lines · Tony Krowiak
Quoted 37 lines · Pierre Morel

Libudev relies on having a subsystem link for non-root devices. To avoid libudev (and potentially other userspace tools) choking on the matrix device let us introduce a vfio_ap bus and with that the vfio_ap bus subsytem, and make the matrix device reside within it.

Doing this we need to suppress the forced link from the matrix device to the vfio_ap driver and we suppress the device_type we do not need anymore.

Since the associated matrix driver is not the vfio_ap driver any more, we have to change the search for the devices on the vfio_ap driver in the function vfio_ap_verify_queue_reserved.

---
  drivers/s390/crypto/vfio_ap_drv.c     | 48
+++++++++++++++++++++++++++++------
  drivers/s390/crypto/vfio_ap_ops.c     |  4 +--
  drivers/s390/crypto/vfio_ap_private.h |  1 +

  3 files changed, 43 insertions(+), 10 deletions(-)
drivers/s390/crypto/vfio_ap_drv.c +0 -4
index 31c6c84..8e45559 100644
@@ -24,10 +24,6 @@ MODULE_LICENSE("GPL v2");
   static struct ap_driver vfio_ap_drv;
-static struct device_type vfio_ap_dev_type = {
-    .name = VFIO_AP_DEV_TYPE_NAME,
-};
-
   struct ap_matrix_dev *matrix_dev;
   /* Only type 10 adapters (CEX4 and later) are supported

drivers/s390/crypto/vfio_ap_drv.c static void vfio_ap_matrix_dev_release(struct · line 64

Quoted 7 lines
drivers/s390/crypto/vfio_ap_drv.c +3 -0
@@ -62,6 +58,27 @@ static void vfio_ap_matrix_dev_release(struct 
 device *dev)
       kfree(matrix_dev);
   }
+static int matrix_bus_match(struct device *dev, struct device_driver 
 *drv)
+{
+    return 1;  

I think we should verify the following:

* dev == matrix_dev->device
* drv == &matrix_driver

Quoted 6 lines
Quoted 4 lines

The model employed is for the matrix device to be a singleton, so I think we should verify that the matrix device and driver defined herein ought to be the only possible choices for a match. Of course, doing so will require some restructuring of this patch.

I think Conny already answered this question.

Not quite :), but I don't think we need any magic in there, as there's only one device and only one driver on that bus. No need to make this more complicated.

Quoted 22 lines
Quoted 13 lines
Quoted 13 lines
drivers/s390/crypto/vfio_ap_drv.c +11 -0
+}
+
+static struct bus_type matrix_bus = {
+    .name = "vfio_ap",
+    .match = &matrix_bus_match,
+};
+
+static int matrix_probe(struct device *dev)
+{
+    return 0;
+}
Quoted 8 lines

drivers/s390/crypto/vfio_ap_drv.c static void vfio_ap_matrix_dev_release(struct device *dev) · line 77

Quoted 3 lines
drivers/s390/crypto/vfio_ap_drv.c +3 -0
+
+static struct device_driver matrix_driver = {
+    .name = "vfio_ap",  

This is the same name used for the original device driver. I think this driver ought to have a different name to avoid confusion. How about vfio_ap_matrix or some other name to differentiate the two.

Quoted 2 lines

I would like too, but changing this will change the path to the mediated device supported type.

Yes, we don't want to change that.

Quoted 9 lines
Quoted 9 lines

drivers/s390/crypto/vfio_ap_drv.c static void vfio_ap_matrix_dev_release(struct device *dev) · line 79

Quoted 5 lines
drivers/s390/crypto/vfio_ap_drv.c +2 -0
+    .bus = &matrix_bus,
+    .probe = matrix_probe,  

I would add:
    .suppress_bind_attrs = true;

Quoted 8 lines
Quoted 4 lines

This will remove the sysfs bind/unbind interfaces. Since there is only one matrix device and it's lifecycle is controlled herein, there is no sense in allowing a root user to bind/unbind it.

OTOH bind/unbind has no impact.
If no one else ask for this I will not change what has already been reviewed by Conny and Christian.

As we only have one driver, it does not really make sense anyway.

maintainer
#10 lore raw
Quoted 60 lines · On Wed, 20 Feb 2019 10:27:31 +0100 Cornelia Huck <[email protected]> wrote:
Quoted 57 lines · On Tue, 19 Feb 2019 22:31:17 +0100 Pierre Morel <[email protected]> wrote:
Quoted 55 lines · Tony Krowiak
Quoted 37 lines · Pierre Morel

Libudev relies on having a subsystem link for non-root devices. To avoid libudev (and potentially other userspace tools) choking on the matrix device let us introduce a vfio_ap bus and with that the vfio_ap bus subsytem, and make the matrix device reside within it.

Doing this we need to suppress the forced link from the matrix device to the vfio_ap driver and we suppress the device_type we do not need anymore.

Since the associated matrix driver is not the vfio_ap driver any more, we have to change the search for the devices on the vfio_ap driver in the function vfio_ap_verify_queue_reserved.

---
  drivers/s390/crypto/vfio_ap_drv.c     | 48
+++++++++++++++++++++++++++++------
  drivers/s390/crypto/vfio_ap_ops.c     |  4 +--
  drivers/s390/crypto/vfio_ap_private.h |  1 +

  3 files changed, 43 insertions(+), 10 deletions(-)
drivers/s390/crypto/vfio_ap_drv.c +0 -4
index 31c6c84..8e45559 100644
@@ -24,10 +24,6 @@ MODULE_LICENSE("GPL v2");
   static struct ap_driver vfio_ap_drv;
-static struct device_type vfio_ap_dev_type = {
-    .name = VFIO_AP_DEV_TYPE_NAME,
-};
-
   struct ap_matrix_dev *matrix_dev;
   /* Only type 10 adapters (CEX4 and later) are supported

drivers/s390/crypto/vfio_ap_drv.c static void vfio_ap_matrix_dev_release(struct · line 64

Quoted 7 lines
drivers/s390/crypto/vfio_ap_drv.c +3 -0
@@ -62,6 +58,27 @@ static void vfio_ap_matrix_dev_release(struct 
 device *dev)
       kfree(matrix_dev);
   }
+static int matrix_bus_match(struct device *dev, struct device_driver 
 *drv)
+{
+    return 1;  

I think we should verify the following:

* dev == matrix_dev->device
* drv == &matrix_driver

The model employed is for the matrix device to be a singleton, so I think we should verify that the matrix device and driver defined herein ought to be the only possible choices for a match. Of course, doing so will require some restructuring of this patch.

Quoted 6 lines
Quoted 1 line

I think Conny already answered this question.

Not quite :), but I don't think we need any magic in there, as there's only one device and only one driver on that bus. No need to make this more complicated.

I agree, no need to complicate this any further.

Quoted 22 lines
Quoted 22 lines
Quoted 21 lines
Quoted 6 lines
drivers/s390/crypto/vfio_ap_drv.c +4 -0
+}
+
+static struct bus_type matrix_bus = {
+    .name = "vfio_ap",

drivers/s390/crypto/vfio_ap_drv.c static void vfio_ap_matrix_dev_release(struct device *dev) · line 77

Quoted 10 lines
drivers/s390/crypto/vfio_ap_drv.c +10 -0
+    .match = &matrix_bus_match,
+};
+
+static int matrix_probe(struct device *dev)
+{
+    return 0;
+}
+
+static struct device_driver matrix_driver = {
+    .name = "vfio_ap",  

This is the same name used for the original device driver. I think this driver ought to have a different name to avoid confusion. How about vfio_ap_matrix or some other name to differentiate the two.

Quoted 5 lines
Quoted 2 lines

I would like too, but changing this will change the path to the mediated device supported type.

Yes, we don't want to change that.

Nod.

Quoted 14 lines
Quoted 14 lines
Quoted 13 lines

drivers/s390/crypto/vfio_ap_drv.c static void vfio_ap_matrix_dev_release(struct device *dev) · line 79

Quoted 5 lines
drivers/s390/crypto/vfio_ap_drv.c +2 -0
+    .bus = &matrix_bus,
+    .probe = matrix_probe,  

I would add:
    .suppress_bind_attrs = true;

This will remove the sysfs bind/unbind interfaces. Since there is only one matrix device and it's lifecycle is controlled herein, there is no sense in allowing a root user to bind/unbind it.

Quoted 6 lines
Quoted 3 lines

OTOH bind/unbind has no impact.
If no one else ask for this I will not change what has already been reviewed by Conny and Christian.

As we only have one driver, it does not really make sense anyway.

I see this as a reason to suppress_bind_attrs. It is much easier than to think about what should happen when one unbinds the matrix device from the vfio_ap driver on the vfio_ap bus. With the code as is it seems to just keep working as if nothing happened.
And /sys/devices/vfio_ap/matrix/mdev_supported_types/ referencing the name of the driver that is already gone sounds a bit weird.

Regards,
Halil

#11 lore raw
Quoted 9 lines · Halil Pasic
Quoted 6 lines · On Wed, 20 Feb 2019 10:27:31 +0100 Cornelia Huck <[email protected]> wrote:
Quoted 3 lines · On Tue, 19 Feb 2019 22:31:17 +0100 Pierre Morel <[email protected]> wrote:
Quoted 2 lines · Tony Krowiak
Quoted 1 line · Pierre Morel

Libudev relies on having a subsystem link for non-root devices. To

...snip...

Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 8 lines

drivers/s390/crypto/vfio_ap_drv.c static void vfio_ap_matrix_dev_release(struct device *dev) · line 77

Quoted 3 lines
drivers/s390/crypto/vfio_ap_drv.c +3 -0
+
+static struct device_driver matrix_driver = {
+    .name = "vfio_ap",

This is the same name used for the original device driver. I think this driver ought to have a different name to avoid confusion. How about vfio_ap_matrix or some other name to differentiate the two.

Quoted 7 lines
Quoted 5 lines
Quoted 2 lines

I would like too, but changing this will change the path to the mediated device supported type.

Yes, we don't want to change that.

Nod.

However if I cannot change the driver name, I can change the bus name to matrix. At least one less "vfio_ap" name

Quoted 29 lines
Quoted 15 lines
Quoted 15 lines
Quoted 14 lines
Quoted 6 lines

+    .bus = &matrix_bus,
+    .probe = matrix_probe,

I would add:
    .suppress_bind_attrs = true;

This will remove the sysfs bind/unbind interfaces. Since there is only one matrix device and it's lifecycle is controlled herein, there is no sense in allowing a root user to bind/unbind it.

Quoted 6 lines
Quoted 3 lines

OTOH bind/unbind has no impact.
If no one else ask for this I will not change what has already been reviewed by Conny and Christian.

As we only have one driver, it does not really make sense anyway.

I see this as a reason to suppress_bind_attrs. It is much easier than to think about what should happen when one unbinds the matrix device from the vfio_ap driver on the vfio_ap bus. With the code as is it seems to just keep working as if nothing happened.
And /sys/devices/vfio_ap/matrix/mdev_supported_types/ referencing the name of the driver that is already gone sounds a bit weird.

Quoted 3 lines

Regards,
Halil

If there is no objection I will do this,
It seems more logical to me too.

Regards,
Pierre

Signature · 2 lines
Pierre Morel
Linux/KVM/QEMU in Böblingen - Germany
#12 lore raw
Quoted 62 lines · Pierre Morel
Quoted 9 lines · Halil Pasic
Quoted 6 lines · On Wed, 20 Feb 2019 10:27:31 +0100 Cornelia Huck <[email protected]> wrote:
Quoted 3 lines · On Tue, 19 Feb 2019 22:31:17 +0100 Pierre Morel <[email protected]> wrote:
Quoted 2 lines · Tony Krowiak
Quoted 1 line · Pierre Morel

Libudev relies on having a subsystem link for non-root devices. To

...snip...

Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 8 lines

drivers/s390/crypto/vfio_ap_drv.c static void vfio_ap_matrix_dev_release(struct device *dev) · line 77

Quoted 3 lines
drivers/s390/crypto/vfio_ap_drv.c +3 -0
+
+static struct device_driver matrix_driver = {
+    .name = "vfio_ap",

This is the same name used for the original device driver. I think this driver ought to have a different name to avoid confusion. How about vfio_ap_matrix or some other name to differentiate the two.

Quoted 7 lines
Quoted 5 lines
Quoted 2 lines

I would like too, but changing this will change the path to the mediated device supported type.

Yes, we don't want to change that.

Nod.

However if I cannot change the driver name, I can change the bus name to matrix. At least one less "vfio_ap" name

Quoted 29 lines
Quoted 15 lines
Quoted 15 lines
Quoted 14 lines

Quoted 2 lines

+    .bus = &matrix_bus,
+    .probe = matrix_probe,

I would add:
      .suppress_bind_attrs = true;

This will remove the sysfs bind/unbind interfaces. Since there is only one matrix device and it's lifecycle is controlled herein, there is no sense in allowing a root user to bind/unbind it.

Quoted 6 lines
Quoted 3 lines

OTOH bind/unbind has no impact.
If no one else ask for this I will not change what has already been reviewed by Conny and Christian.

As we only have one driver, it does not really make sense anyway.

I see this as a reason to suppress_bind_attrs. It is much easier than to think about what should happen when one unbinds the matrix device from the vfio_ap driver on the vfio_ap bus. With the code as is it seems to just keep working as if nothing happened.
And /sys/devices/vfio_ap/matrix/mdev_supported_types/ referencing the name of the driver that is already gone sounds a bit weird.

Quoted 6 lines
Quoted 3 lines

Regards,
Halil

If there is no objection I will do this,
It seems more logical to me too.

Go ahead and send this as v3?

#13 lore raw
Quoted 65 lines · Christian Borntraeger
Quoted 62 lines · Pierre Morel
Quoted 9 lines · Halil Pasic
Quoted 6 lines · On Wed, 20 Feb 2019 10:27:31 +0100 Cornelia Huck <[email protected]> wrote:
Quoted 3 lines · On Tue, 19 Feb 2019 22:31:17 +0100 Pierre Morel <[email protected]> wrote:
Quoted 2 lines · Tony Krowiak
Quoted 1 line · Pierre Morel

Libudev relies on having a subsystem link for non-root devices. To

...snip...

Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 8 lines

Review

Quoted 3 lines
(quoted diff) +3 -0
+
+static struct device_driver matrix_driver = {
+    .name = "vfio_ap",

This is the same name used for the original device driver. I think this driver ought to have a different name to avoid confusion. How about vfio_ap_matrix or some other name to differentiate the two.

Quoted 7 lines
Quoted 5 lines
Quoted 2 lines

I would like too, but changing this will change the path to the mediated device supported type.

Yes, we don't want to change that.

Nod.

However if I cannot change the driver name, I can change the bus name to matrix. At least one less "vfio_ap" name

Quoted 29 lines
Quoted 15 lines
Quoted 15 lines
Quoted 14 lines
Quoted 6 lines

+    .bus = &matrix_bus,
+    .probe = matrix_probe,

I would add:
      .suppress_bind_attrs = true;

This will remove the sysfs bind/unbind interfaces. Since there is only one matrix device and it's lifecycle is controlled herein, there is no sense in allowing a root user to bind/unbind it.

Quoted 6 lines
Quoted 3 lines

OTOH bind/unbind has no impact.
If no one else ask for this I will not change what has already been reviewed by Conny and Christian.

As we only have one driver, it does not really make sense anyway.

I see this as a reason to suppress_bind_attrs. It is much easier than to think about what should happen when one unbinds the matrix device from the vfio_ap driver on the vfio_ap bus. With the code as is it seems to just keep working as if nothing happened.
And /sys/devices/vfio_ap/matrix/mdev_supported_types/ referencing the name of the driver that is already gone sounds a bit weird.

Quoted 9 lines
Quoted 6 lines
Quoted 3 lines

Regards,
Halil

If there is no objection I will do this,
It seems more logical to me too.

Go ahead and send this as v3?

OK
CC stable ?

Signature · 2 lines
Pierre Morel
Linux/KVM/QEMU in Böblingen - Germany
#14 lore raw
Quoted 3 lines · Pierre Morel
Quoted 2 lines · Christian Borntraeger

[..]

Quoted 5 lines
Quoted 2 lines

Go ahead and send this as v3?

OK
CC stable ?

Yes, also add a "Fixes:" tag.

#15 lore raw
Quoted 6 lines · Christian Borntraeger
Quoted 3 lines · Pierre Morel
Quoted 2 lines · Christian Borntraeger

[..]

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Go ahead and send this as v3?

OK
CC stable ?

Yes, also add a "Fixes:" tag.

Yes, I will not forget this time :)

Thanks
Pierre

Signature · 2 lines
Pierre Morel
Linux/KVM/QEMU in Böblingen - Germany
#16 lore raw
In reply to Pierre Morel
Quoted 142 lines · Pierre Morel

Libudev relies on having a subsystem link for non-root devices. To avoid libudev (and potentially other userspace tools) choking on the matrix device let us introduce a vfio_ap bus and with that the vfio_ap bus subsytem, and make the matrix device reside within it.

Doing this we need to suppress the forced link from the matrix device to the vfio_ap driver and we suppress the device_type we do not need anymore.

Since the associated matrix driver is not the vfio_ap driver any more, we have to change the search for the devices on the vfio_ap driver in the function vfio_ap_verify_queue_reserved.

---
 drivers/s390/crypto/vfio_ap_drv.c     | 48 +++++++++++++++++++++++++++++------
 drivers/s390/crypto/vfio_ap_ops.c     |  4 +--
 drivers/s390/crypto/vfio_ap_private.h |  1 +
 3 files changed, 43 insertions(+), 10 deletions(-)
drivers/s390/crypto/vfio_ap_drv.c +40 -8
index 31c6c84..8e45559 100644
@@ -24,10 +24,6 @@ MODULE_LICENSE("GPL v2");
 
 static struct ap_driver vfio_ap_drv;
 
-static struct device_type vfio_ap_dev_type = {
-	.name = VFIO_AP_DEV_TYPE_NAME,
-};
-
 struct ap_matrix_dev *matrix_dev;
 
 /* Only type 10 adapters (CEX4 and later) are supported
@@ -62,6 +58,27 @@ static void vfio_ap_matrix_dev_release(struct device *dev)
 	kfree(matrix_dev);
 }
 
+static int matrix_bus_match(struct device *dev, struct device_driver *drv)
+{
+	return 1;
+}
+
+static struct bus_type matrix_bus = {
+	.name = "vfio_ap",
+	.match = &matrix_bus_match,
+};
+
+static int matrix_probe(struct device *dev)
+{
+	return 0;
+}
+
+static struct device_driver matrix_driver = {
+	.name = "vfio_ap",
+	.bus = &matrix_bus,
+	.probe = matrix_probe,
+};
+
 static int vfio_ap_matrix_dev_create(void)
 {
 	int ret;
@@ -71,6 +88,10 @@ static int vfio_ap_matrix_dev_create(void)
 	if (IS_ERR(root_device))
 		return PTR_ERR(root_device);
 
+	ret = bus_register(&matrix_bus);
+	if (ret)
+		goto bus_register_err;
+
 	matrix_dev = kzalloc(sizeof(*matrix_dev), GFP_KERNEL);
 	if (!matrix_dev) {
 		ret = -ENOMEM;
@@ -87,30 +108,41 @@ static int vfio_ap_matrix_dev_create(void)
 	mutex_init(&matrix_dev->lock);
 	INIT_LIST_HEAD(&matrix_dev->mdev_list);
 
-	matrix_dev->device.type = &vfio_ap_dev_type;
 	dev_set_name(&matrix_dev->device, "%s", VFIO_AP_DEV_NAME);
 	matrix_dev->device.parent = root_device;
+	matrix_dev->device.bus = &matrix_bus;
 	matrix_dev->device.release = vfio_ap_matrix_dev_release;
-	matrix_dev->device.driver = &vfio_ap_drv.driver;
+	matrix_dev->vfio_ap_drv = &vfio_ap_drv;
 
 	ret = device_register(&matrix_dev->device);
 	if (ret)
 		goto matrix_reg_err;
 
+	ret = driver_register(&matrix_driver);
+	if (ret)
+		goto matrix_drv_err;
+
 	return 0;
 
+matrix_drv_err:
+	device_unregister(&matrix_dev->device);
 matrix_reg_err:
 	put_device(&matrix_dev->device);
 matrix_alloc_err:
+	bus_unregister(&matrix_bus);
+bus_register_err:
 	root_device_unregister(root_device);
-
 	return ret;
 }
 
 static void vfio_ap_matrix_dev_destroy(void)
 {
+	struct device *root_device = matrix_dev->device.parent;
+
+	driver_unregister(&matrix_driver);
 	device_unregister(&matrix_dev->device);
-	root_device_unregister(matrix_dev->device.parent);
+	bus_unregister(&matrix_bus);
+	root_device_unregister(root_device);
 }
 
 static int __init vfio_ap_init(void)
drivers/s390/crypto/vfio_ap_ops.c +2 -2
index 272ef42..900b9cf 100644
@@ -198,8 +198,8 @@ static int vfio_ap_verify_queue_reserved(unsigned long *apid,
 	qres.apqi = apqi;
 	qres.reserved = false;
 
-	ret = driver_for_each_device(matrix_dev->device.driver, NULL, &qres,
-				     vfio_ap_has_queue);
+	ret = driver_for_each_device(&matrix_dev->vfio_ap_drv->driver, NULL,
+				     &qres, vfio_ap_has_queue);
 	if (ret)
 		return ret;
 
drivers/s390/crypto/vfio_ap_private.h +0 -0
index 5675492..76b7f98 100644

drivers/s390/crypto/vfio_ap_private.h struct ap_matrix_dev { · line 46

Quoted 7 lines
drivers/s390/crypto/vfio_ap_private.h +1 -0
@@ -40,6 +40,7 @@ struct ap_matrix_dev {
 	struct ap_config_info info;
 	struct list_head mdev_list;
 	struct mutex lock;
+	struct ap_driver  *vfio_ap_drv;
 };
 
 extern struct ap_matrix_dev *matrix_dev;

You are introducing a new bus just for a user space application which is unable to deal with a device directly attached to the root of devices ? So you are fixing kernel code because of disability of userspace code. Hm, you are switching root cause and effect. However, not my job.

Why do you need this dummy bus ? Did you evaluate using a "class" subsystem instead ? This is very common and my assumption is that libudev is able to handle this. I am using a "zcrypt" class for providing additional zcrypt device nodes and this works fine together with udev. I would avoid the introduction and maintenance of bus code at any cost.

Btw. having a look onto the naming ... the module is named "vfio_ap", the driver is named "vfio_ap", the bus is named "vfio_ap", the root bus device is named "vfio_ap" ... a bunch of vfio_aps naming different things.

regards
Harald Freudenberger

#17 lore raw
Quoted 149 lines · Harald Freudenberger
Quoted 148 lines · Pierre Morel

Libudev relies on having a subsystem link for non-root devices. To avoid libudev (and potentially other userspace tools) choking on the matrix device let us introduce a vfio_ap bus and with that the vfio_ap bus subsytem, and make the matrix device reside within it.

Doing this we need to suppress the forced link from the matrix device to the vfio_ap driver and we suppress the device_type we do not need anymore.

Since the associated matrix driver is not the vfio_ap driver any more, we have to change the search for the devices on the vfio_ap driver in the function vfio_ap_verify_queue_reserved.

---
 drivers/s390/crypto/vfio_ap_drv.c     | 48 +++++++++++++++++++++++++++++------
 drivers/s390/crypto/vfio_ap_ops.c     |  4 +--
 drivers/s390/crypto/vfio_ap_private.h |  1 +
 3 files changed, 43 insertions(+), 10 deletions(-)
drivers/s390/crypto/vfio_ap_drv.c +40 -8
index 31c6c84..8e45559 100644
@@ -24,10 +24,6 @@ MODULE_LICENSE("GPL v2");
 
 static struct ap_driver vfio_ap_drv;
 
-static struct device_type vfio_ap_dev_type = {
-	.name = VFIO_AP_DEV_TYPE_NAME,
-};
-
 struct ap_matrix_dev *matrix_dev;
 
 /* Only type 10 adapters (CEX4 and later) are supported
@@ -62,6 +58,27 @@ static void vfio_ap_matrix_dev_release(struct device *dev)
 	kfree(matrix_dev);
 }
 
+static int matrix_bus_match(struct device *dev, struct device_driver *drv)
+{
+	return 1;
+}
+
+static struct bus_type matrix_bus = {
+	.name = "vfio_ap",
+	.match = &matrix_bus_match,
+};
+
+static int matrix_probe(struct device *dev)
+{
+	return 0;
+}
+
+static struct device_driver matrix_driver = {
+	.name = "vfio_ap",
+	.bus = &matrix_bus,
+	.probe = matrix_probe,
+};
+
 static int vfio_ap_matrix_dev_create(void)
 {
 	int ret;
@@ -71,6 +88,10 @@ static int vfio_ap_matrix_dev_create(void)
 	if (IS_ERR(root_device))
 		return PTR_ERR(root_device);
 
+	ret = bus_register(&matrix_bus);
+	if (ret)
+		goto bus_register_err;
+
 	matrix_dev = kzalloc(sizeof(*matrix_dev), GFP_KERNEL);
 	if (!matrix_dev) {
 		ret = -ENOMEM;
@@ -87,30 +108,41 @@ static int vfio_ap_matrix_dev_create(void)
 	mutex_init(&matrix_dev->lock);
 	INIT_LIST_HEAD(&matrix_dev->mdev_list);
 
-	matrix_dev->device.type = &vfio_ap_dev_type;
 	dev_set_name(&matrix_dev->device, "%s", VFIO_AP_DEV_NAME);
 	matrix_dev->device.parent = root_device;
+	matrix_dev->device.bus = &matrix_bus;
 	matrix_dev->device.release = vfio_ap_matrix_dev_release;
-	matrix_dev->device.driver = &vfio_ap_drv.driver;
+	matrix_dev->vfio_ap_drv = &vfio_ap_drv;
 
 	ret = device_register(&matrix_dev->device);
 	if (ret)
 		goto matrix_reg_err;
 
+	ret = driver_register(&matrix_driver);
+	if (ret)
+		goto matrix_drv_err;
+
 	return 0;
 
+matrix_drv_err:
+	device_unregister(&matrix_dev->device);
 matrix_reg_err:
 	put_device(&matrix_dev->device);
 matrix_alloc_err:
+	bus_unregister(&matrix_bus);
+bus_register_err:
 	root_device_unregister(root_device);
-
 	return ret;
 }
 
 static void vfio_ap_matrix_dev_destroy(void)
 {
+	struct device *root_device = matrix_dev->device.parent;
+
+	driver_unregister(&matrix_driver);
 	device_unregister(&matrix_dev->device);
-	root_device_unregister(matrix_dev->device.parent);
+	bus_unregister(&matrix_bus);
+	root_device_unregister(root_device);
 }
 
 static int __init vfio_ap_init(void)
drivers/s390/crypto/vfio_ap_ops.c +2 -2
index 272ef42..900b9cf 100644
@@ -198,8 +198,8 @@ static int vfio_ap_verify_queue_reserved(unsigned long *apid,
 	qres.apqi = apqi;
 	qres.reserved = false;
 
-	ret = driver_for_each_device(matrix_dev->device.driver, NULL, &qres,
-				     vfio_ap_has_queue);
+	ret = driver_for_each_device(&matrix_dev->vfio_ap_drv->driver, NULL,
+				     &qres, vfio_ap_has_queue);
 	if (ret)
 		return ret;
 
drivers/s390/crypto/vfio_ap_private.h +1 -0
index 5675492..76b7f98 100644
@@ -40,6 +40,7 @@ struct ap_matrix_dev {
 	struct ap_config_info info;
 	struct list_head mdev_list;
 	struct mutex lock;
+	struct ap_driver  *vfio_ap_drv;
 };
 
Quoted 6 lines

drivers/s390/crypto/vfio_ap_private.h struct ap_matrix_dev { · line 46

Quoted 1 line
drivers/s390/crypto/vfio_ap_private.h +0 -0
 extern struct ap_matrix_dev *matrix_dev;

You are introducing a new bus just for a user space application which is unable to deal with a device directly attached to the root of devices ? So you are fixing kernel code because of disability of userspace code. Hm, you are switching root cause and effect. However, not my job.

the kernel rule is pretty simple. If userspace breaks due to a kernel change fix the kernel.

Quoted 6 lines

Why do you need this dummy bus ? Did you evaluate using a "class" subsystem instead ? This is very common and my assumption is that libudev is able to handle this. I am using a "zcrypt" class for providing additional zcrypt device nodes and this works fine together with udev. I would avoid the introduction and maintenance of bus code at any cost.

The class variant sounds promising. Pierre what do you think?

Quoted 8 lines

Btw. having a look onto the naming ... the module is named "vfio_ap", the driver is named "vfio_ap", the bus is named "vfio_ap", the root bus device is named "vfio_ap" ... a bunch of vfio_aps naming different things.

regards
Harald Freudenberger

#18 lore raw
Quoted 159 lines · Christian Borntraeger
Quoted 149 lines · Harald Freudenberger
Quoted 148 lines · Pierre Morel

Libudev relies on having a subsystem link for non-root devices. To avoid libudev (and potentially other userspace tools) choking on the matrix device let us introduce a vfio_ap bus and with that the vfio_ap bus subsytem, and make the matrix device reside within it.

Doing this we need to suppress the forced link from the matrix device to the vfio_ap driver and we suppress the device_type we do not need anymore.

Since the associated matrix driver is not the vfio_ap driver any more, we have to change the search for the devices on the vfio_ap driver in the function vfio_ap_verify_queue_reserved.

---
 drivers/s390/crypto/vfio_ap_drv.c     | 48 +++++++++++++++++++++++++++++------
 drivers/s390/crypto/vfio_ap_ops.c     |  4 +--
 drivers/s390/crypto/vfio_ap_private.h |  1 +
 3 files changed, 43 insertions(+), 10 deletions(-)
drivers/s390/crypto/vfio_ap_drv.c +40 -8
index 31c6c84..8e45559 100644
@@ -24,10 +24,6 @@ MODULE_LICENSE("GPL v2");
 
 static struct ap_driver vfio_ap_drv;
 
-static struct device_type vfio_ap_dev_type = {
-	.name = VFIO_AP_DEV_TYPE_NAME,
-};
-
 struct ap_matrix_dev *matrix_dev;
 
 /* Only type 10 adapters (CEX4 and later) are supported
@@ -62,6 +58,27 @@ static void vfio_ap_matrix_dev_release(struct device *dev)
 	kfree(matrix_dev);
 }
 
+static int matrix_bus_match(struct device *dev, struct device_driver *drv)
+{
+	return 1;
+}
+
+static struct bus_type matrix_bus = {
+	.name = "vfio_ap",
+	.match = &matrix_bus_match,
+};
+
+static int matrix_probe(struct device *dev)
+{
+	return 0;
+}
+
+static struct device_driver matrix_driver = {
+	.name = "vfio_ap",
+	.bus = &matrix_bus,
+	.probe = matrix_probe,
+};
+
 static int vfio_ap_matrix_dev_create(void)
 {
 	int ret;
@@ -71,6 +88,10 @@ static int vfio_ap_matrix_dev_create(void)
 	if (IS_ERR(root_device))
 		return PTR_ERR(root_device);
 
+	ret = bus_register(&matrix_bus);
+	if (ret)
+		goto bus_register_err;
+
 	matrix_dev = kzalloc(sizeof(*matrix_dev), GFP_KERNEL);
 	if (!matrix_dev) {
 		ret = -ENOMEM;
@@ -87,30 +108,41 @@ static int vfio_ap_matrix_dev_create(void)
 	mutex_init(&matrix_dev->lock);
 	INIT_LIST_HEAD(&matrix_dev->mdev_list);
 
-	matrix_dev->device.type = &vfio_ap_dev_type;
 	dev_set_name(&matrix_dev->device, "%s", VFIO_AP_DEV_NAME);
 	matrix_dev->device.parent = root_device;
+	matrix_dev->device.bus = &matrix_bus;
 	matrix_dev->device.release = vfio_ap_matrix_dev_release;
-	matrix_dev->device.driver = &vfio_ap_drv.driver;
+	matrix_dev->vfio_ap_drv = &vfio_ap_drv;
 
 	ret = device_register(&matrix_dev->device);
 	if (ret)
 		goto matrix_reg_err;
 
+	ret = driver_register(&matrix_driver);
+	if (ret)
+		goto matrix_drv_err;
+
 	return 0;
 
+matrix_drv_err:
+	device_unregister(&matrix_dev->device);
 matrix_reg_err:
 	put_device(&matrix_dev->device);
 matrix_alloc_err:
+	bus_unregister(&matrix_bus);
+bus_register_err:
 	root_device_unregister(root_device);
-
 	return ret;
 }
 
 static void vfio_ap_matrix_dev_destroy(void)
 {
+	struct device *root_device = matrix_dev->device.parent;
+
+	driver_unregister(&matrix_driver);
 	device_unregister(&matrix_dev->device);
-	root_device_unregister(matrix_dev->device.parent);
+	bus_unregister(&matrix_bus);
+	root_device_unregister(root_device);
 }
 
 static int __init vfio_ap_init(void)
drivers/s390/crypto/vfio_ap_ops.c +2 -2
index 272ef42..900b9cf 100644
@@ -198,8 +198,8 @@ static int vfio_ap_verify_queue_reserved(unsigned long *apid,
 	qres.apqi = apqi;
 	qres.reserved = false;
 
-	ret = driver_for_each_device(matrix_dev->device.driver, NULL, &qres,
-				     vfio_ap_has_queue);
+	ret = driver_for_each_device(&matrix_dev->vfio_ap_drv->driver, NULL,
+				     &qres, vfio_ap_has_queue);
 	if (ret)
 		return ret;
 
drivers/s390/crypto/vfio_ap_private.h +1 -0
index 5675492..76b7f98 100644
@@ -40,6 +40,7 @@ struct ap_matrix_dev {
 	struct ap_config_info info;
 	struct list_head mdev_list;
 	struct mutex lock;
+	struct ap_driver  *vfio_ap_drv;
 };
 
Quoted 5 lines

drivers/s390/crypto/vfio_ap_private.h struct ap_matrix_dev { · line 46

Quoted 1 line
drivers/s390/crypto/vfio_ap_private.h +0 -0
 extern struct ap_matrix_dev *matrix_dev;

You are introducing a new bus just for a user space application which is unable to deal with a device directly attached to the root of devices ? So you are fixing kernel code because of disability of userspace code. Hm, you are switching root cause and effect. However, not my job.

the kernel rule is pretty simple. If userspace breaks due to a kernel change fix the kernel.

Quoted 6 lines
Quoted 5 lines

Why do you need this dummy bus ? Did you evaluate using a "class" subsystem instead ? This is very common and my assumption is that libudev is able to handle this. I am using a "zcrypt" class for providing additional zcrypt device nodes and this works fine together with udev. I would avoid the introduction and maintenance of bus code at any cost.

The class variant sounds promising. Pierre what do you think?

I checked that. My additional zcrypt devices (you can easily create one with just echo "blubber" >/sys/class/zcrypt/create) and then have a look to the new device entry in sysfs at /sys/devices/virtual/zcrypt/blubber). Maybe you need to alloc device numbers for this - I am not sure if it is sufficient to device_register() without a device number. However, have a look into zcrpyt_api.c zcdn_init() and zcdn_create().
With a new class, you can also remove the root device needed as an anchor for the old and for the bus code.

Quoted 7 lines
Quoted 7 lines

Btw. having a look onto the naming ... the module is named "vfio_ap", the driver is named "vfio_ap", the bus is named "vfio_ap", the root bus device is named "vfio_ap" ... a bunch of vfio_aps naming different things.

regards
Harald Freudenberger

#19 lore raw
In reply to Christian Borntraeger
Quoted 163 lines · Christian Borntraeger
Quoted 149 lines · Harald Freudenberger
Quoted 148 lines · Pierre Morel

Libudev relies on having a subsystem link for non-root devices. To avoid libudev (and potentially other userspace tools) choking on the matrix device let us introduce a vfio_ap bus and with that the vfio_ap bus subsytem, and make the matrix device reside within it.

Doing this we need to suppress the forced link from the matrix device to the vfio_ap driver and we suppress the device_type we do not need anymore.

Since the associated matrix driver is not the vfio_ap driver any more, we have to change the search for the devices on the vfio_ap driver in the function vfio_ap_verify_queue_reserved.

---
  drivers/s390/crypto/vfio_ap_drv.c     | 48 +++++++++++++++++++++++++++++------
  drivers/s390/crypto/vfio_ap_ops.c     |  4 +--
  drivers/s390/crypto/vfio_ap_private.h |  1 +
  3 files changed, 43 insertions(+), 10 deletions(-)
drivers/s390/crypto/vfio_ap_drv.c +40 -8
index 31c6c84..8e45559 100644
@@ -24,10 +24,6 @@ MODULE_LICENSE("GPL v2");
 
  static struct ap_driver vfio_ap_drv;
 
-static struct device_type vfio_ap_dev_type = {
-	.name = VFIO_AP_DEV_TYPE_NAME,
-};
-
  struct ap_matrix_dev *matrix_dev;
 
  /* Only type 10 adapters (CEX4 and later) are supported
@@ -62,6 +58,27 @@ static void vfio_ap_matrix_dev_release(struct device *dev)
  	kfree(matrix_dev);
  }
 
+static int matrix_bus_match(struct device *dev, struct device_driver *drv)
+{
+	return 1;
+}
+
+static struct bus_type matrix_bus = {
+	.name = "vfio_ap",
+	.match = &matrix_bus_match,
+};
+
+static int matrix_probe(struct device *dev)
+{
+	return 0;
+}
+
+static struct device_driver matrix_driver = {
+	.name = "vfio_ap",
+	.bus = &matrix_bus,
+	.probe = matrix_probe,
+};
+
  static int vfio_ap_matrix_dev_create(void)
  {
  	int ret;
@@ -71,6 +88,10 @@ static int vfio_ap_matrix_dev_create(void)
  	if (IS_ERR(root_device))
  		return PTR_ERR(root_device);
 
+	ret = bus_register(&matrix_bus);
+	if (ret)
+		goto bus_register_err;
+
  	matrix_dev = kzalloc(sizeof(*matrix_dev), GFP_KERNEL);
  	if (!matrix_dev) {
  		ret = -ENOMEM;
@@ -87,30 +108,41 @@ static int vfio_ap_matrix_dev_create(void)
  	mutex_init(&matrix_dev->lock);
  	INIT_LIST_HEAD(&matrix_dev->mdev_list);
 
-	matrix_dev->device.type = &vfio_ap_dev_type;
  	dev_set_name(&matrix_dev->device, "%s", VFIO_AP_DEV_NAME);
  	matrix_dev->device.parent = root_device;
+	matrix_dev->device.bus = &matrix_bus;
  	matrix_dev->device.release = vfio_ap_matrix_dev_release;
-	matrix_dev->device.driver = &vfio_ap_drv.driver;
+	matrix_dev->vfio_ap_drv = &vfio_ap_drv;
 
  	ret = device_register(&matrix_dev->device);
  	if (ret)
  		goto matrix_reg_err;
 
+	ret = driver_register(&matrix_driver);
+	if (ret)
+		goto matrix_drv_err;
+
  	return 0;
 
+matrix_drv_err:
+	device_unregister(&matrix_dev->device);
  matrix_reg_err:
  	put_device(&matrix_dev->device);
  matrix_alloc_err:
+	bus_unregister(&matrix_bus);
+bus_register_err:
  	root_device_unregister(root_device);
-
  	return ret;
  }
 
  static void vfio_ap_matrix_dev_destroy(void)
  {
+	struct device *root_device = matrix_dev->device.parent;
+
+	driver_unregister(&matrix_driver);
  	device_unregister(&matrix_dev->device);
-	root_device_unregister(matrix_dev->device.parent);
+	bus_unregister(&matrix_bus);
+	root_device_unregister(root_device);
  }
 
  static int __init vfio_ap_init(void)
drivers/s390/crypto/vfio_ap_ops.c +2 -2
index 272ef42..900b9cf 100644
@@ -198,8 +198,8 @@ static int vfio_ap_verify_queue_reserved(unsigned long *apid,
  	qres.apqi = apqi;
  	qres.reserved = false;
 
-	ret = driver_for_each_device(matrix_dev->device.driver, NULL, &qres,
-				     vfio_ap_has_queue);
+	ret = driver_for_each_device(&matrix_dev->vfio_ap_drv->driver, NULL,
+				     &qres, vfio_ap_has_queue);
  	if (ret)
  		return ret;
 
drivers/s390/crypto/vfio_ap_private.h +1 -0
index 5675492..76b7f98 100644
@@ -40,6 +40,7 @@ struct ap_matrix_dev {
  	struct ap_config_info info;
  	struct list_head mdev_list;
  	struct mutex lock;
+	struct ap_driver  *vfio_ap_drv;
  };
 
Quoted 6 lines

drivers/s390/crypto/vfio_ap_private.h struct ap_matrix_dev { · line 46

Quoted 1 line
drivers/s390/crypto/vfio_ap_private.h +0 -0
  extern struct ap_matrix_dev *matrix_dev;

You are introducing a new bus just for a user space application which is unable to deal with a device directly attached to the root of devices ? So you are fixing kernel code because of disability of userspace code. Hm, you are switching root cause and effect. However, not my job.

the kernel rule is pretty simple. If userspace breaks due to a kernel change fix the kernel.

Quoted 1 line
Quoted 7 lines
Quoted 5 lines

Why do you need this dummy bus ? Did you evaluate using a "class" subsystem instead ? This is very common and my assumption is that libudev is able to handle this. I am using a "zcrypt" class for providing additional zcrypt device nodes and this works fine together with udev. I would avoid the introduction and maintenance of bus code at any cost.

The class variant sounds promising. Pierre what do you think?

AFAIU the device driver model, the struct class is higher level representation of the device used to provide the informations to create devices in /dev. The device and the class are associated with the call to device_create which send the event.

However the device passed to device_create must hang on a bus and be managed by a driver.

I do not see the interest of having a class if we do not plan to create a device in /dev. We could add one in prevision, but it is not a replacement for the bus.

Regards,
Pierre

Signature · 2 lines
Pierre Morel
Linux/KVM/QEMU in Böblingen - Germany