[PATCH] VFIO: remove dead notifier code

[PATCH] VFIO: remove dead notifier code

[Paolo Bonzini]

group->notifier is dead code. VFIO initializes it and checks it for emptiness on
teardown, but nobody ever registers on it or triggers it.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 Documentation/arch/s390/vfio-ap.rst | 18 ++++++------------
 drivers/vfio/group.c                |  3 ---
 drivers/vfio/vfio.h                 |  1 -
 drivers/vfio/vfio_iommu_type1.c     |  1 -
 4 files changed, 6 insertions(+), 17 deletions(-)

diff --git a/Documentation/arch/s390/vfio-ap.rst b/Documentation/arch/s390/vfio-ap.rst
index eba1991fbdba..2de0560bd70c 100644
--- a/Documentation/arch/s390/vfio-ap.rst
+++ b/Documentation/arch/s390/vfio-ap.rst
@@ -431,17 +431,14 @@ matrix device.
 * callback interfaces
 
   open_device:
-    The vfio_ap driver uses this callback to register a
-    VFIO_GROUP_NOTIFY_SET_KVM notifier callback function for the matrix mdev
-    devices. The open_device callback is invoked by userspace to connect the
-    VFIO iommu group for the matrix mdev device to the MDEV bus. Access to the
-    KVM structure used to configure the KVM guest is provided via this callback.
-    The KVM structure, is used to configure the guest's access to the AP matrix
-    defined via the vfio_ap mediated device's sysfs attribute files.
+    the open_device callback is invoked by userspace to connect the
+    VFIO iommu group for the matrix mdev device to the MDEV bus.  The
+    callback retrieves the KVM structure used to configure the KVM guest
+    and configures the guest's access to the AP matrix defined via the
+    vfio_ap mediated device's sysfs attribute files.
 
   close_device:
-    unregisters the VFIO_GROUP_NOTIFY_SET_KVM notifier callback function for the
-    matrix mdev device and deconfigures the guest's AP matrix.
+    this callback deconfigures the guest's AP matrix.
 
   ioctl:
     this callback handles the VFIO_DEVICE_GET_INFO and VFIO_DEVICE_RESET ioctls
@@ -449,9 +444,8 @@ matrix device.
 
 Configure the guest's AP resources
 ----------------------------------
-Configuring the AP resources for a KVM guest will be performed when the
-VFIO_GROUP_NOTIFY_SET_KVM notifier callback is invoked. The notifier
-function is called when userspace connects to KVM. The guest's AP resources are
+Configuring the AP resources for a KVM guest will be performed at the
+time of ``open_device`` and ``close_device``. The guest's AP resources are
 configured via its APCB by:
 
 * Setting the bits in the APM corresponding to the APIDs assigned to the
diff --git a/drivers/vfio/group.c b/drivers/vfio/group.c
index 4f15016d2a5f..731dd0dead0d 100644
--- a/drivers/vfio/group.c
+++ b/drivers/vfio/group.c
@@ -456,7 +456,6 @@ static int vfio_group_fops_release(struct inode *inode, struct file *filep)
 	 * Device FDs hold a group file reference, therefore the group release
 	 * is only called when there are no open devices.
 	 */
-	WARN_ON(group->notifier.head);
 	if (group->container)
 		vfio_group_detach_container(group);
 	if (group->iommufd) {
@@ -541,7 +540,6 @@ static struct vfio_group *vfio_group_alloc(struct iommu_group *iommu_group,
 	/* put in vfio_group_release() */
 	iommu_group_ref_get(iommu_group);
 	group->type = type;
-	BLOCKING_INIT_NOTIFIER_HEAD(&group->notifier);
 
 	return group;
 }
@@ -720,7 +718,6 @@ void vfio_device_remove_group(struct vfio_device *device)
 	 * properly hold the group reference.
 	 */
 	WARN_ON(!list_empty(&group->device_list));
-	WARN_ON(group->notifier.head);
 
 	/*
 	 * Revoke all users of group->iommu_group. At this point we know there
diff --git a/drivers/vfio/vfio.h b/drivers/vfio/vfio.h
index 50128da18bca..0854f3fa1a22 100644
--- a/drivers/vfio/vfio.h
+++ b/drivers/vfio/vfio.h
@@ -90,7 +90,6 @@ struct vfio_group {
 	struct mutex			group_lock;
 	struct kvm			*kvm;
 	struct file			*opened_file;
-	struct blocking_notifier_head	notifier;
 	struct iommufd_ctx		*iommufd;
 	spinlock_t			kvm_ref_lock;
 	unsigned int			cdev_device_open_cnt;
diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
index 03cefdf99a4a..c8151ba54de3 100644
--- a/drivers/vfio/vfio_iommu_type1.c
+++ b/drivers/vfio/vfio_iommu_type1.c
@@ -36,7 +36,6 @@
 #include <linux/uaccess.h>
 #include <linux/vfio.h>
 #include <linux/workqueue.h>
-#include <linux/notifier.h>
 #include <linux/mm_inline.h>
 #include <linux/overflow.h>
 #include "vfio.h"
-- 
2.53.0

Re: [PATCH] VFIO: remove dead notifier code

[Anthony Krowiak]

On 4/7/26 1:59 PM, Paolo Bonzini wrote:
> group->notifier is dead code. VFIO initializes it and checks it for emptiness on
> teardown, but nobody ever registers on it or triggers it.
>
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

There are two minor review comments about capitalization at the 
beginning of
sentences in the vfio-ap.rst document in the Documentation/arch/s390
directory. Other than that:

Acked-by: Anthony Krowiak <akrowiak@linux.ibm.com>

> ---
>   Documentation/arch/s390/vfio-ap.rst | 18 ++++++------------
>   drivers/vfio/group.c                |  3 ---
>   drivers/vfio/vfio.h                 |  1 -
>   drivers/vfio/vfio_iommu_type1.c     |  1 -
>   4 files changed, 6 insertions(+), 17 deletions(-)
>
> diff --git a/Documentation/arch/s390/vfio-ap.rst b/Documentation/arch/s390/vfio-ap.rst
> index eba1991fbdba..2de0560bd70c 100644
> --- a/Documentation/arch/s390/vfio-ap.rst
> +++ b/Documentation/arch/s390/vfio-ap.rst
> @@ -431,17 +431,14 @@ matrix device.
>   * callback interfaces
>   
>     open_device:
> -    The vfio_ap driver uses this callback to register a
> -    VFIO_GROUP_NOTIFY_SET_KVM notifier callback function for the matrix mdev
> -    devices. The open_device callback is invoked by userspace to connect the
> -    VFIO iommu group for the matrix mdev device to the MDEV bus. Access to the
> -    KVM structure used to configure the KVM guest is provided via this callback.
> -    The KVM structure, is used to configure the guest's access to the AP matrix
> -    defined via the vfio_ap mediated device's sysfs attribute files.
> +    the open_device callback is invoked by userspace to connect the

Capitalization:
s/the open_device callback/The open device callback/

> +    VFIO iommu group for the matrix mdev device to the MDEV bus.  The
> +    callback retrieves the KVM structure used to configure the KVM guest
> +    and configures the guest's access to the AP matrix defined via the
> +    vfio_ap mediated device's sysfs attribute files.
>   
>     close_device:
> -    unregisters the VFIO_GROUP_NOTIFY_SET_KVM notifier callback function for the
> -    matrix mdev device and deconfigures the guest's AP matrix.
> +    this callback deconfigures the guest's AP matrix.

Capitalization:
s/this callback/This callback/

>   
>     ioctl:
>       this callback handles the VFIO_DEVICE_GET_INFO and VFIO_DEVICE_RESET ioctls
> @@ -449,9 +444,8 @@ matrix device.
>   
>   Configure the guest's AP resources
>   ----------------------------------
> -Configuring the AP resources for a KVM guest will be performed when the
> -VFIO_GROUP_NOTIFY_SET_KVM notifier callback is invoked. The notifier
> -function is called when userspace connects to KVM. The guest's AP resources are
> +Configuring the AP resources for a KVM guest will be performed at the
> +time of ``open_device`` and ``close_device``. The guest's AP resources are
>   configured via its APCB by:
>   
>   * Setting the bits in the APM corresponding to the APIDs assigned to the
> diff --git a/drivers/vfio/group.c b/drivers/vfio/group.c
> index 4f15016d2a5f..731dd0dead0d 100644
> --- a/drivers/vfio/group.c
> +++ b/drivers/vfio/group.c
> @@ -456,7 +456,6 @@ static int vfio_group_fops_release(struct inode *inode, struct file *filep)
>   	 * Device FDs hold a group file reference, therefore the group release
>   	 * is only called when there are no open devices.
>   	 */
> -	WARN_ON(group->notifier.head);
>   	if (group->container)
>   		vfio_group_detach_container(group);
>   	if (group->iommufd) {
> @@ -541,7 +540,6 @@ static struct vfio_group *vfio_group_alloc(struct iommu_group *iommu_group,
>   	/* put in vfio_group_release() */
>   	iommu_group_ref_get(iommu_group);
>   	group->type = type;
> -	BLOCKING_INIT_NOTIFIER_HEAD(&group->notifier);
>   
>   	return group;
>   }
> @@ -720,7 +718,6 @@ void vfio_device_remove_group(struct vfio_device *device)
>   	 * properly hold the group reference.
>   	 */
>   	WARN_ON(!list_empty(&group->device_list));
> -	WARN_ON(group->notifier.head);
>   
>   	/*
>   	 * Revoke all users of group->iommu_group. At this point we know there
> diff --git a/drivers/vfio/vfio.h b/drivers/vfio/vfio.h
> index 50128da18bca..0854f3fa1a22 100644
> --- a/drivers/vfio/vfio.h
> +++ b/drivers/vfio/vfio.h
> @@ -90,7 +90,6 @@ struct vfio_group {
>   	struct mutex			group_lock;
>   	struct kvm			*kvm;
>   	struct file			*opened_file;
> -	struct blocking_notifier_head	notifier;
>   	struct iommufd_ctx		*iommufd;
>   	spinlock_t			kvm_ref_lock;
>   	unsigned int			cdev_device_open_cnt;
> diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
> index 03cefdf99a4a..c8151ba54de3 100644
> --- a/drivers/vfio/vfio_iommu_type1.c
> +++ b/drivers/vfio/vfio_iommu_type1.c
> @@ -36,7 +36,6 @@
>   #include <linux/uaccess.h>
>   #include <linux/vfio.h>
>   #include <linux/workqueue.h>
> -#include <linux/notifier.h>
>   #include <linux/mm_inline.h>
>   #include <linux/overflow.h>
>   #include "vfio.h"

Re: [PATCH] VFIO: remove dead notifier code

[Paolo Bonzini]

On Wed, Apr 8, 2026 at 4:35 PM Anthony Krowiak <akrowiak@linux.ibm.com> wrote:
>
>
> >     open_device:
> > -    The vfio_ap driver uses this callback to register a
> > -    VFIO_GROUP_NOTIFY_SET_KVM notifier callback function for the matrix mdev
> > -    devices. The open_device callback is invoked by userspace to connect the
> > -    VFIO iommu group for the matrix mdev device to the MDEV bus. Access to the
> > -    KVM structure used to configure the KVM guest is provided via this callback.
> > -    The KVM structure, is used to configure the guest's access to the AP matrix
> > -    defined via the vfio_ap mediated device's sysfs attribute files.
> > +    the open_device callback is invoked by userspace to connect the
>
> Capitalization:
> s/the open_device callback/The open device callback/
>
> > +    VFIO iommu group for the matrix mdev device to the MDEV bus.  The
> > +    callback retrieves the KVM structure used to configure the KVM guest
> > +    and configures the guest's access to the AP matrix defined via the
> > +    vfio_ap mediated device's sysfs attribute files.
> >
> >     close_device:
> > -    unregisters the VFIO_GROUP_NOTIFY_SET_KVM notifier callback function for the
> > -    matrix mdev device and deconfigures the guest's AP matrix.
> > +    this callback deconfigures the guest's AP matrix.
>
> Capitalization:
> s/this callback/This callback/

Heh, I had to pick one or the other (close_device previously used
lowercase) and chose the wrong one. :)

Thanks for reviewing the s390 docs, I wasn't sure about the new text was enough.

Paolo

Re: [PATCH] VFIO: remove dead notifier code

[Paolo Bonzini]

On Wed, Apr 8, 2026 at 4:36 PM Paolo Bonzini <pbonzini@redhat.com> wrote:
>
> On Wed, Apr 8, 2026 at 4:35 PM Anthony Krowiak <akrowiak@linux.ibm.com> wrote:
> >
> >
> > >     open_device:
> > > -    The vfio_ap driver uses this callback to register a
> > > -    VFIO_GROUP_NOTIFY_SET_KVM notifier callback function for the matrix mdev
> > > -    devices. The open_device callback is invoked by userspace to connect the
> > > -    VFIO iommu group for the matrix mdev device to the MDEV bus. Access to the
> > > -    KVM structure used to configure the KVM guest is provided via this callback.
> > > -    The KVM structure, is used to configure the guest's access to the AP matrix
> > > -    defined via the vfio_ap mediated device's sysfs attribute files.
> > > +    the open_device callback is invoked by userspace to connect the
> >
> > Capitalization:
> > s/the open_device callback/The open device callback/
> >
> > Capitalization:
> > s/this callback/This callback/
>
> Heh, I had to pick one or the other (close_device previously used
> lowercase) and chose the wrong one. :)

I was going to make the change, but then noticed that lowercase is
more common in the file:

  create:
    allocates the ap_matrix_mdev structure used by the vfio_ap driver to:

  remove:
    deallocates the vfio_ap mediated device's ap_matrix_mdev structure.
    This will be allowed only if a running guest is not using the mdev.

  ioctl:
    this callback handles the VFIO_DEVICE_GET_INFO and VFIO_DEVICE_RESET
    ioctls defined by the vfio framework.

What do you think?

Paolo

Re: [PATCH] VFIO: remove dead notifier code

[Anthony Krowiak]

On 4/8/26 6:00 PM, Paolo Bonzini wrote:
> On Wed, Apr 8, 2026 at 4:36 PM Paolo Bonzini <pbonzini@redhat.com> wrote:
>> On Wed, Apr 8, 2026 at 4:35 PM Anthony Krowiak <akrowiak@linux.ibm.com> wrote:
>>>
>>>>      open_device:
>>>> -    The vfio_ap driver uses this callback to register a
>>>> -    VFIO_GROUP_NOTIFY_SET_KVM notifier callback function for the matrix mdev
>>>> -    devices. The open_device callback is invoked by userspace to connect the
>>>> -    VFIO iommu group for the matrix mdev device to the MDEV bus. Access to the
>>>> -    KVM structure used to configure the KVM guest is provided via this callback.
>>>> -    The KVM structure, is used to configure the guest's access to the AP matrix
>>>> -    defined via the vfio_ap mediated device's sysfs attribute files.
>>>> +    the open_device callback is invoked by userspace to connect the
>>> Capitalization:
>>> s/the open_device callback/The open device callback/
>>>
>>> Capitalization:
>>> s/this callback/This callback/
>> Heh, I had to pick one or the other (close_device previously used
>> lowercase) and chose the wrong one. :)
> I was going to make the change, but then noticed that lowercase is
> more common in the file:
>
>    create:
>      allocates the ap_matrix_mdev structure used by the vfio_ap driver to:
>
>    remove:
>      deallocates the vfio_ap mediated device's ap_matrix_mdev structure.
>      This will be allowed only if a running guest is not using the mdev.
>
>    ioctl:
>      this callback handles the VFIO_DEVICE_GET_INFO and VFIO_DEVICE_RESET
>      ioctls defined by the vfio framework.
>
> What do you think?

It's not critical. If that's the case, leave it lower case

>
> Paolo
>

Re: [PATCH] VFIO: remove dead notifier code

[Jason Gunthorpe]

On Tue, Apr 07, 2026 at 07:59:34PM +0200, Paolo Bonzini wrote:
> group->notifier is dead code. VFIO initializes it and checks it for emptiness on
> teardown, but nobody ever registers on it or triggers it.
> 
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
>  Documentation/arch/s390/vfio-ap.rst | 18 ++++++------------
>  drivers/vfio/group.c                |  3 ---
>  drivers/vfio/vfio.h                 |  1 -
>  drivers/vfio/vfio_iommu_type1.c     |  1 -
>  4 files changed, 6 insertions(+), 17 deletions(-)

 Fixes: 421cfe6596f6 ("vfio: remove VFIO_GROUP_NOTIFY_SET_KVM")

?

Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>

Jason

Re: [PATCH] VFIO: remove dead notifier code

[Alex Williamson]

On Fri, 10 Apr 2026 15:10:06 -0300
Jason Gunthorpe <jgg@ziepe.ca> wrote:

> On Tue, Apr 07, 2026 at 07:59:34PM +0200, Paolo Bonzini wrote:
> > group->notifier is dead code. VFIO initializes it and checks it for emptiness on
> > teardown, but nobody ever registers on it or triggers it.
> > 
> > Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> > ---
> >  Documentation/arch/s390/vfio-ap.rst | 18 ++++++------------
> >  drivers/vfio/group.c                |  3 ---
> >  drivers/vfio/vfio.h                 |  1 -
> >  drivers/vfio/vfio_iommu_type1.c     |  1 -
> >  4 files changed, 6 insertions(+), 17 deletions(-)  
> 
>  Fixes: 421cfe6596f6 ("vfio: remove VFIO_GROUP_NOTIFY_SET_KVM")
> 
> ?

Technically yes, but is it really worth having stable pick this up?
Seems not.  Thanks,

Alex