Re: [Patch 1/1] vfio: Move "device->open_count--" out of group_rwsem in vfio_device_open()

[Yi Liu <yi.l.liu@intel.com>]

On 2022/6/21 10:49, Tian, Kevin wrote:
>> From: Liu, Yi L <yi.l.liu@intel.com>
>> Sent: Tuesday, June 21, 2022 9:32 AM
>>
>> On 2022/6/21 04:13, Matthew Rosato wrote:
>>> On 6/20/22 4:54 AM, Yi Liu wrote:
>>>> No need to protect open_count with group_rwsem
>>>>
>>>> Fixes: 421cfe6596f6 ("vfio: remove VFIO_GROUP_NOTIFY_SET_KVM")
>>>>
>>>> cc: Matthew Rosato <mjrosato@linux.ibm.com>
>>>> cc: Jason Gunthorpe <jgg@nvidia.com>
>>>> Signed-off-by: Yi Liu <yi.l.liu@intel.com>
>>>
>>> Seems pretty harmless as-is, but you are correct group_rwsem can be
>> dropped
>>> earlier; we do not protect the count with group_rwsem elsewhere (see
>>> vfio_device_fops_release as a comparison, where we already drop
>> group_rwsem
>>> before open_count--)
>>
>> yes. this is exactly how I found it. Normally, I compare the err handling
>> path with the release function to see if they are aligned. :-)
> 
> In this case we don't need a FIX tag. It's kind of optimization.

ok.

>>
>>> FWIW, this change now also drops group_rswem before setting device-
>>> kvm =
>>> NULL, but that's also OK (again, just like vfio_device_fops_release) --
>>> While 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.
>>
>> yes. set device->kvm to be NULL has no need to hold group_rwsem. BTW. I
>> also doubt whether the device->ops->open_device(device) and
>> device->ops->close_device(device) should be protected by group_rwsem or
>> not. seems not, right? group_rwsem protects the fields under vfio_group.
>> For the open_device/close_device() device->dev_set->lock is enough. Maybe
>> another nit fix.
>>
> 
> group->rwsem is to protect device->group->kvm from being changed
> by vfio_file_set_kvm() before it is copied to device->kvm.

yes. this is why vfio_device_open() holds the read lock of group_rwsem 
around the device->group->kvm copy. However, for the open_device(), 
callback, I don't think it is necessary to be protected by the group_rwsem
lock.

-- 
Regards,
Yi Liu