vfio: Move "device->open_count--" out of group_rwsem in vfio_device_open()
Subsystems (2)
PATCH v2 1/1 vfio: Move "device->open_count--" out of group_rwsem in vfio_device_open()
We do not protect the vfio_device::open_count with group_rwsem elsewhere (see vfio_device_fops_release as a comparison, where we already drop group_rwsem before open_count--). So move the group_rwsem unlock prior to open_count--.
This change now also drops group_rswem before setting device->kvm = NULL, but that's also OK (again, just like vfio_device_fops_release). The setting of device->kvm before open_device is technically done while holding the group_rwsem, this is done to protect the group kvm value we are copying from, and we should not be relying on that to protect the contents of device->kvm; instead we assume this value will not change until after the device is closed and while under the dev_set->lock.
--- drivers/vfio/vfio.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
---
v2:
- Remove Fixes tag (Kevin)
- Add detailed description in commit message (Matthew, Jason)
- Fix patch format (Jason)
- Add r-b from Matthew
v1:
https://lore.kernel.org/kvm/[email protected]/
---
drivers/vfio/vfio.c +1 -1
@@ -1146,10 +1146,10 @@ static struct file *vfio_device_open(struct vfio_device *device)if (device->open_count ==1 && device->ops->close_device) device->ops->close_device (device); err_undo_count: +up_read (&device->group->group_rwsem); device->open_count--;if (device->open_count ==0 && device->kvm) device->kvm =NULL ; -up_read (&device->group->group_rwsem);mutex_unlock (&device->dev_set->lock);module_put (device->dev->driver->owner); err_unassign_container:
Signature · 1 line
2.27.0
Quoted 19 lines
From: Liu, Yi L <[email protected]>
Sent: Monday, June 27, 2022 3:41 PMWe do not protect the
vfio_device::open_countwithgroup_rwsemelsewhere (seevfio_device_fops_releaseas a comparison, where we already dropgroup_rwsem
before open_count--). So move thegroup_rwsemunlock prior to open_count--.This change now also drops
group_rswembefore settingdevice->kvm= NULL, but that's also OK (again, just likevfio_device_fops_release). The setting ofdevice->kvmbeforeopen_deviceis technically done while holding thegroup_rwsem, this is done to protect the group kvm value we are copying from,
and we should not be relying on that to protect the contents ofdevice->kvm; instead we assume this value will not change until after the device is closed and while under thedev_set->lock.
Quoted 4 lines
Cc Matthew Rosato <[email protected]>Cc Jason Gunthorpe <[email protected]>Signed-off-by Yi Liu <[email protected]>Reviewed-by Matthew Rosato <[email protected]>
Alex Williamson
AppliedQuoted 12 lines · On Mon, 27 Jun 2022 00:41:19 -0700 Yi Liu <[email protected]> wrote:
We do not protect the
vfio_device::open_countwithgroup_rwsemelsewhere (seevfio_device_fops_releaseas a comparison, where we already dropgroup_rwsembefore open_count--). So move thegroup_rwsemunlock prior to open_count--.This change now also drops
group_rswembefore settingdevice->kvm= NULL, but that's also OK (again, just likevfio_device_fops_release). The setting ofdevice->kvmbeforeopen_deviceis technically done while holding thegroup_rwsem, this is done to protect the group kvm value we are copying from, and we should not be relying on that to protect the contents ofdevice->kvm; instead we assume this value will not change until after the device is closed and while under thedev_set->lock.
Quoted 8 lines
Cc Matthew Rosato <[email protected]>Cc Jason Gunthorpe <[email protected]>Signed-off-by Yi Liu <[email protected]>Reviewed-by Matthew Rosato <[email protected]>--- drivers/vfio/vfio.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)---
Applied to vfio next branch for v5.20. Thanks,
Alex
Quoted 26 lines
v2:
- Remove Fixes tag (Kevin)
- Add detailed description in commit message (Matthew, Jason)
- Fix patch format (Jason)
- Add r-b from Matthewv1:
https://lore.kernel.org/kvm/[email protected]/
---drivers/vfio/vfio.c +1
-1@@ -1146,10 +1146,10 @@ static struct file *vfio_device_open(struct vfio_device *device)if (device->open_count ==1 && device->ops->close_device) device->ops->close_device (device); err_undo_count: +up_read (&device->group->group_rwsem); device->open_count--;if (device->open_count ==0 && device->kvm) device->kvm =NULL ; -up_read (&device->group->group_rwsem);mutex_unlock (&device->dev_set->lock);module_put (device->dev->driver->owner); err_unassign_container: