Add support for dynamic UAR usage

This series enables the user space driver to ask for a dynamic UAR allocation and uses its internal BF (i.e. blue-flame) registers as part of QP creation.

The motivation behind this functionality is to let the user space driver share BF register between few QPs or to set a dedicated BF for a given QP.

The above gives an option to user space applications to utilize its resources and performance according to its needs.

For example:
An application can improve its performance by having a dedicated BF for collection of QPs that are running from the same thread and drop the need to take a lock as part of its post send flow. This is not enabled today as the kernel driver assigned BF registers in some FIFO logic and as such QPs from different threads may share same BF and as such a lock when accessing this BF will be needed.

The series consists from 3 patches:
The first patch extends the alloc context flow to be prepared for working with dynamic UAR allocations.
The second patch exposes the option to dynamic allocates a UAR. The third patch uses the given BF register from the user area as part of QP creation instead of using some arbitrary BF register.

The patches are available in the git repository at:
git.kernel.org/pub/scm/linux/kernel/git/leon/linux-rdma.git tags/rdma-next-2017-12-24-5

Thanks
---------------------------------------

Yishai Hadas (3):
IB/mlx5: Extend UAR stuff to support dynamic allocation IB/mlx5: Expose dynamic mmap allocation
IB/mlx5: Enable QP creation with a given blue flame index

 drivers/infiniband/hw/mlx5/main.c    | 141 ++++++++++++++++++++++++++++-------
 drivers/infiniband/hw/mlx5/mlx5_ib.h |  13 +++-
 drivers/infiniband/hw/mlx5/qp.c      |  55 ++++++++++----
 include/linux/mlx5/device.h          |   2 +
 include/linux/mlx5/driver.h          |   3 +
 include/uapi/rdma/mlx5-abi.h         |   5 +-
 6 files changed, 177 insertions(+), 42 deletions(-)

--
2.15.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html