On Sun, 2018-09-16 at 20:48 +0300, Leon Romanovsky wrote:
> From: Leon Romanovsky <leonro@xxxxxxxxxxxx>
>
> From Jason,
>
> Restricting ODP to use only a single process and mm doesn't really make
> any sense, the data structure work to fix that is fairly straightforward
> and allows ODP to also stop using tgid as well.
>
> This series also cleans up a few things in the ODP code along the way
> to make it easier to make the datastructure change, and fixes a bunch
> of locking race conditions and other mistakes that syzkaller and lockdep
> finds in this code.
>
> Thanks
>
> Jason Gunthorpe (9):
> RDMA/umem: Use ib_umem_odp in all function signatures connected to ODP
> RDMA/umem: Make ib_umem_odp into a sub structure of ib_umem
> RDMA/umem: Get rid of struct ib_umem.odp_data
> RDMA/umem: Move all the ODP related stuff out of ucontext and into
> per_mm
> RDMA/umem: Use umem->owning_mm inside ODP
> RDMA/umem: Get rid of per_mm->notifier_count
> RDMA/umem: Handle a half-complete start/end sequence
> RDMA/umem: Avoid synchronize_srcu in the ODP MR destruction path
> RDMA/uverbs: Get rid of ucontext->tgid
Series looks nice! Applied to for-next.
--
Doug Ledford <dledford@xxxxxxxxxx>
GPG KeyID: B826A3330E572FDD
Key fingerprint = AE6B 1BDA 122B 23B4 265B 1274 B826 A333 0E57 2FDD
Attachment:
signature.asc
Description: This is a digitally signed message part
