Re: [PATCH v4 09/19] vfio/pci: Accept device fd for hot reset

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Sun, Feb 26, 2023 at 08:59:01AM +0000, Liu, Yi L wrote:
> > From: Jason Gunthorpe <jgg@xxxxxxxxxx>
> > Sent: Friday, February 24, 2023 10:36 AM
> > 
> > On Fri, Feb 24, 2023 at 02:21:33AM +0000, Tian, Kevin wrote:
> > 
> > > Yi, while you are incorporating this change please also update the
> > > uapi header. Rename 'group_fds[]' to 'fds[]' and add comment to
> > > explain that it could be an array of group fds or a single iommufd.
> > 
> > Upon reflection we can probably make it even simpler and just have a 0
> > length fd array mean to use the iommufd the vfio_device is already
> > associated with
> > 
> > And the check for correctness can be simplified to simply see if each
> > vfio_device in the dev_set is attached to the same iommufd ctx
> > pointer instead of searching the xarray.
> 
> Sorry, it appears to me the below concern is not solved as above logic
> still requires userspace to open and bind devices to the same iommufd.
> 
> "
>   > Say a scenario where group happens to overlap with devset. Let's say
>   > two devices in the group/devset.
>   > 
>   > An existing deployment assigns only dev1 to Qemu. In this case dev1
>   > is resettable via group fd given dev2 cannot be opened by another
>   > user.
> "

You solve this by checking for a 0 open count as already discussed.

Jason



[Index of Archives]     [KVM ARM]     [KVM ia64]     [KVM ppc]     [Virtualization Tools]     [Spice Development]     [Libvirt]     [Libvirt Users]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite Questions]     [Linux Kernel]     [Linux SCSI]     [XFree86]

  Powered by Linux