xen summit 2023 design session: Upstreaming PCIe passthrough on ARM

[Luca Fancellu <Luca.Fancellu@xxxxxxx>]

- Bertrand: we are talking about the work for PCI passthrough on ARM (and PVH 
[Stefano])
- Stefano: the problem is that it's taking a lot of time to merge the serie, 
how to make progress?
  Stewart is allocated on that, original work from epam and arm, the feature 
works, is not perfect, but it's testable.
- Jan: main problem is the lock
[...]
- Roger: the last status is looking good, maybe just a couple of deadlock
- Stefano: Some problem comes also from the fact that contributor and 
maintainer should sync
- Jan: Roger: [technical discussion about locks in pci structures, lock 
inversion]
- Jan: We should not introduce too many locks
- Stefano: Bertrand: agree, we should extend current locks
- Roger: Maybe per domain locks is the best one for now
- Bertrand: I agree on that, partially, even if it is not ideal, but for now it 
could be ok.
- Roger: I would like the serie rebased
- Stefano: yes I can get the people working on that to rebase and sync between 
them. when I was reviewing the locking, it was difficult to review without the 
other pieces.
- Roger: There is a slow turn around in the mailing list
- Bertrand: Can Stewart help on the lock problem and code changes? 
- Stefano: I have 2 actions, sync with epam to have the rebase, and sync with 
Stewart to address in a fast pace the comments coming from the mailing list
 - Bertrand: we need to manage to have the locking. We need to continue review 
also in code freeze, because we can review even if we don't commit. So our goal 
is 4.19.
 - Jan: I usually tend to not send patches apart from bugfixes on code freeze, 
but in this case it can be ok because this feature is stuck from a long time.


Sorry but I wasn’t able to take more refined notes.

Cheers,
Luca