Re: [PATCH 0/4] Fix hwpt lifetime and detach bugs

[Jason Gunthorpe <jgg@nvidia.com>]

On Wed, Feb 22, 2023 at 02:14:42AM +0000, Tian, Kevin wrote:
> > From: Jason Gunthorpe <jgg@nvidia.com>
> > Sent: Thursday, February 16, 2023 9:37 AM
> > 
> > On Mon, Feb 13, 2023 at 02:02:41PM -0400, Jason Gunthorpe wrote:
> > > As discussed here is a small series to address the confusing lifetime
> > > scheme for the hwpt. This was some leftover from prior rework that was
> > > never fully cleaned up.
> > >
> > > Make it clear that the ioas and ioas->hwpt_list are associated with the
> > > hwpt during creation and never changed until it is destroyed. A
> > > post-finalize hwpt with a positive reference count is always valid for
> > > device attachment.
> > >
> > > This should simplify the nesting and replace series.
> > >
> > > Jason Gunthorpe (4):
> > >   iommufd: Do not add the same hwpt to the ioas->hwpt_list twice
> > 
> > I've applied this patch to the tree
> > 
> > And my v2 of this series is here now:
> > 
> > https://github.com/jgunthorpe/linux/commits/iommufd_hwpt
> > 
> > Which is completely redone
> > 
> > I will try to make the patches to do replace and alloc hwpt on top of
> > this, I think it fixes all the locking and consistency problems.
> > 
> 
> with group now explicitly tracked in iommufd presumably iommufd
> can continue to use iommu_attach_group() instead of switching to
> iommu_attach_device() as previously planned?

Yes, it seems necessary.

The only other option I could think of was to expose more of the
core's iommu_group struct for owner usage.

Jason