Re: [PATCH 1/3] vfio: Introduce vma ops registration and notifier

On Thu, 18 Feb 2021 19:04:04 -0400
Jason Gunthorpe [off-list ref] wrote:

On Thu, Feb 18, 2021 at 02:56:06PM -0700, Alex Williamson wrote:

quoted

Looks pretty slick.  I won't claim it's fully gelled in my head yet,
but AIUI you're creating these inodes on your new pseudo fs and
associating it via the actual user fd via the f_mapping pointer, which
allows multiple fds to associate and address space back to this inode
when you want to call unmap_mapping_range().    

Yes, from what I can tell all the fs/inode stuff is just mandatory
overhead to get a unique address_space pointer, as that is the only
thing this is actually using.

I have to check the mmap flow more carefully, I recall pointing to a
existing race here with Daniel, but the general idea should hold.

quoted

That clarifies from the previous email how we'd store the inode on
the vfio_device without introducing yet another tracking list for
device fds.  

Right, you can tell from the vma what inode it is for, and the inode
can tell you if it is a VFIO VMA or not, so no tracking lists needed
at all.

Seems to be a nice cleanup for vfio as well, more testing and analysis
required, but here are a few (4) wip commits that re-implement the
current vma zapping scheme following your example: 

https://github.com/awilliam/linux-vfio/commits/vfio-unmap-mapping-range

Thanks,
Alex