Support bind a qp with counter. If counter is null then bind the qp to the
default counter. Different QP state has different operation:

- RESET: Set the counter field so that it will take effective during
  RST2INIT change;
- RTS: Issue an RTS2RTS change to update the QP counter;
- Other: Set the counter field and mark the counter_pending flag, when QP
  is moved to RTS state and this flag is set, then issue an RTS2RTS
  modification to update the counter.
Signed-off-by: Mark Zhang <markz@mellanox.com>
Reviewed-by: Majd Dibbiny <majd@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

In auto mode all QPs belong to one category are bind automatically to a
single counter set. Currently only "qp type" is supported.

In this mode the qp counter is set in RST2INIT modification, and when a qp
is destroyed the counter is unbound.
Signed-off-by: Mark Zhang <markz@mellanox.com>
Reviewed-by: Majd Dibbiny <majd@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

Add an API to support set/clear per-port auto mode.
Signed-off-by: Mark Zhang <markz@mellanox.com>
Reviewed-by: Majd Dibbiny <majd@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

Remove is_visible_in_pid_ns() from nldev.c and make it as a restrack API,
so that it can be taken advantage by other parts like counter.
Signed-off-by: Mark Zhang <markz@mellanox.com>
Reviewed-by: Majd Dibbiny <majd@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

Add rdma_restrack_attach_task() which is able to attach a task other then
"current" to a resource.
Signed-off-by: Mark Zhang <markz@mellanox.com>
Reviewed-by: Majd Dibbiny <majd@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

Introduce statistic counter as a new resource. It allows a user to monitor
specific objects (e.g., QPs) by binding to a counter.

In some cases a user counter resource is created with task other then
"current", because its creation is done as part of rdmatool call.
Signed-off-by: Mark Zhang <markz@mellanox.com>
Reviewed-by: Majd Dibbiny <majd@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

From git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux

Required for dependencies in the next patches.

* mlx5-next:
  net/mlx5: Add rts2rts_qp_counters_set_id field in hca cap
  net/mlx5: Properly name the generic WQE control field
  net/mlx5: Introduce TLS TX offload hardware bits and structures
  net/mlx5: Refactor mlx5_esw_query_functions for modularity
  net/mlx5: E-Switch prepare functions change handler to be modular
  net/mlx5: Introduce and use mlx5_eswitch_get_total_vports()

Add rts2rts_qp_counters_set_id field in hca cap so that RTS2RTS
qp modification can be used to change the counter of a QP.
Signed-off-by: Mark Zhang <markz@mellanox.com>
Reviewed-by: Majd Dibbiny <majd@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>

Make admin commands id easier to distinguish by using relevant bits from
the producer counter.
This allows us to differentiate admin commands with the same producer
index (happens after admin queue overlap), which is helpful when
debugging.
Signed-off-by: Daniel Kranzdorf <dkkranzd@amazon.com>
Reviewed-by: Firas JahJah <firasj@amazon.com>
Reviewed-by: Yossi Leybovich <sleybo@amazon.com>
Signed-off-by: Gal Pressman <galpress@amazon.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

Despite failure in ipoib_dev_init() we continue with initialization flow
and creation of child device. It causes to the situation where this child
device is added too early to parent device list.

Change the logic, so in case of failure we properly return error from
ipoib_dev_init() and add child only in success path.

Fixes: eaeb3984 ("IB/ipoib: Move init code to ndo_init")
Signed-off-by: Valentine Fatiev <valentinef@mellanox.com>
Reviewed-by: Feras Daoud <ferasda@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

There is no need in custom memory zeroing, because it can be done
by using kzalloc from the beginning.
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

The variable ret is being initialized with a value that is never read and
it is being updated later with a new value. The initialization is
redundant and can be removed.

Addresses-Coverity: ("Unused value")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

The patch below wasn't fully tested for all combinations of module and
configs, and causes a compile failure:

WARNING: modpost: missing MODULE_LICENSE() in drivers/infiniband/hw/hns/hns_roce_ah.o
see include/linux/module.h for more information
WARNING: modpost: missing MODULE_LICENSE() in drivers/infiniband/hw/hns/hns_roce_alloc.o
see include/linux/module.h for more information
WARNING: modpost: missing MODULE_LICENSE() in drivers/infiniband/hw/hns/hns_roce_cmd.o
see include/linux/module.h for more information
WARNING: modpost: missing MODULE_LICENSE() in drivers/infiniband/hw/hns/hns_roce_cq.o
see include/linux/module.h for more information
WARNING: modpost: missing MODULE_LICENSE() in drivers/infiniband/hw/hns/hns_roce_db.o
see include/linux/module.h for more information
WARNING: modpost: missing MODULE_LICENSE() in drivers/infiniband/hw/hns/hns_roce_hem.o
see include/linux/module.h for more information
WARNING: modpost: missing MODULE_LICENSE() in drivers/infiniband/hw/hns/hns_roce_mr.o
see include/linux/module.h for more information
WARNING: modpost: missing MODULE_LICENSE() in drivers/infiniband/hw/hns/hns_roce_pd.o
see include/linux/module.h for more information
WARNING: modpost: missing MODULE_LICENSE() in drivers/infiniband/hw/hns/hns_roce_qp.o
see include/linux/module.h for more information
WARNING: modpost: missing MODULE_LICENSE() in drivers/infiniband/hw/hns/hns_roce_restrack.o
see include/linux/module.h for more information
WARNING: modpost: missing MODULE_LICENSE() in drivers/infiniband/hw/hns/hns_roce_srq.o
see include/linux/module.h for more information
ERROR: "hns_roce_bitmap_cleanup" [drivers/infiniband/hw/hns/hns_roce_srq.ko] undefined!
ERROR: "hns_roce_bitmap_init" [drivers/infiniband/hw/hns/hns_roce_srq.ko] undefined!
ERROR: "hns_roce_free_cmd_mailbox" [drivers/infiniband/hw/hns/hns_roce_srq.ko] undefined!
ERROR: "hns_roce_alloc_cmd_mailbox" [drivers/infiniband/hw/hns/hns_roce_srq.ko] undefined!
ERROR: "hns_roce_table_get" [drivers/infiniband/hw/hns/hns_roce_srq.ko] undefined!
ERROR: "hns_roce_bitmap_alloc" [drivers/infiniband/hw/hns/hns_roce_srq.ko] undefined!
ERROR: "hns_roce_table_find" [drivers/infiniband/hw/hns/hns_roce_srq.ko] undefined!

The fix is to put the module sub components in the right line.

Fixes: e9816ddf ("RDMA/hns: Cleanup unnecessary exported symbols")
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Xi Wang <wangxi11@huawei.com>
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

No need any more to hold mlx5_core_dev on the devx_object, it can be
accessed from ib_dev.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

Add DEVX support for CQ events by creating and destroying the CQ via
mlx5_core and set an handler to manage its completions.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

Implement DEVX dispatching event by looking up for the applicable
subscriptions for the reported event and using their target fd to
signal/set the event.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

Enable subscription for device events over DEVX.

Each subscription is added to the two level xarray data structure
according to its event number and the DEVX object information in case was
given with the given target fd.

Those events will be reported over the given fd once will occur.
Downstream patches will mange the dispatching to any subscription.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

Register DEVX with with mlx5_core to get async events.  This will enable
to dispatch the applicable events to its consumers in down stream patches.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

Introduce MLX5_IB_OBJECT_DEVX_ASYNC_EVENT_FD and its initial
implementation.

This object is from type class FD and will be used to read DEVX
async events.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

A generic WQE control field is used for different purposes
in different cases.
Use union to allow using the proper name in each case.
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

Add TLS offload related IFC structs, layouts and enumerations.
Signed-off-by: Eran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

Functions change event output data size changes when functions other
than VFs will be enabled in HCA CAP.
With current API, multiple callers needs to align, calculate accurate
size of the output data depending on number on non VF functions enabled
in the device.
Instead of duplicating such math at multiple places, refactor
mlx5_esw_query_functions() to return raw output allocated by itself.

Caller must free the allocated memory using kvfree() as described in the
function comment section.
This hides calcuation within mlx5_esw_query_functions() and provides
simpler API.
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

Eswitch function change handler will service multiple type of events for
VFs and non VF functions update.
Hence, introduce and use the helper function
esw_vfs_changed_event_handler() for handling change in num VFs to improve
the code readability.
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

Instead MLX5_TOTAL_VPORTS, use mlx5_eswitch_get_total_vports().
mlx5_eswitch_get_total_vports() in subsequent patch accounts for SF
vports as well.
Expanding MLX5_TOTAL_VPORTS macro would require exposing SF internals to
more generic vport.h header file. Such exposure is not desired.
Hence a mlx5_eswitch_get_total_vports() is introduced.

Given that mlx5_eswitch_get_total_vports() API wants to work on const
mlx5_core_dev*, change its helper functions also to accept const *dev.
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

From git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux

Required for dependencies in the next patches.

Resolved the conflicts:
 - esw_destroy_offloads_acl_tables() use the newer mlx5_esw_for_all_vports()
   version
 - esw_offloads_steering_init() drop the cap test
 - esw_offloads_init() drop the extra function arguments

* branch 'mlx5-next': (39 commits)
  net/mlx5: Expose device definitions for object events
  net/mlx5: Report EQE data upon CQ completion
  net/mlx5: Report a CQ error event only when a handler was set
  net/mlx5: mlx5_core_create_cq() enhancements
  net/mlx5: Expose the API to register for ANY event
  net/mlx5: Use event mask based on device capabilities
  net/mlx5: Fix mlx5_core_destroy_cq() error flow
  net/mlx5: E-Switch, Handle UC address change in switchdev mode
  net/mlx5: E-Switch, Consider host PF for inline mode and vlan pop
  net/mlx5: E-Switch, Use iterator for vlan and min-inline setups
  net/mlx5: E-Switch, Reg/unreg function changed event at correct stage
  net/mlx5: E-Switch, Consolidate eswitch function number of VFs
  net/mlx5: E-Switch, Refactor eswitch SR-IOV interface
  net/mlx5: Handle host PF vport mac/guid for ECPF
  net/mlx5: E-Switch, Use correct flags when configuring vlan
  net/mlx5: Reduce dependency on enabled_vfs counter and num_vfs
  net/mlx5: Don't handle VF func change if host PF is disabled
  net/mlx5: Limit scope of mlx5_get_next_phys_dev() to PCI PF devices
  net/mlx5: Move pci status reg access mutex to mlx5_pci_init
  net/mlx5: Rename mlx5_pci_dev_type to mlx5_coredev_type
  ...
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

Currently during dual port IB device registration in below code flow,

ib_register_device()
  ib_device_register_sysfs()
    ib_setup_port_attrs()
      add_port()
        get_counter_table()
          get_perf_mad()
            process_mad()
              mlx5_ib_process_mad()

mlx5_ib_process_mad() fails on 2nd port when both the ports are not fully
setup at the device level (because 2nd port is unaffiliated).

As a result, get_perf_mad() registers different PMA counter group for 1st
and 2nd port, namely pma_counter_ext and pma_counter. However both ports
have the same capability and counter offsets.

Due to this when counters are read by the user via sysfs in below code
flow, counters are queried from wrong location from the device mainly from
PPCNT instead of VPORT counters.

show_pma_counter()
  get_perf_mad()
    process_mad()
      mlx5_ib_process_mad()
        process_pma_cmd()

This shows all zero counters for 2nd port.

To overcome this, process_pma_cmd() is invoked, and when unaffiliated port
is not yet setup during device registration phase, make the query on the
first port.  while at it, only process_pma_cmd() needs to work on the
native port number and underlying mdev, so shift the get, put calls to
where its needed inside process_pma_cmd().

Fixes: 212f2a87 ("IB/mlx5: Route MADs for dual port RoCE")
Signed-off-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Daniel Jurgens <danielj@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

Expose an extra device definitions for objects events.

It includes: object_type values for legacy objects and generic data
header for any other object.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>

Report EQE data upon CQ completion to let upper layers use this data.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>

Report a CQ error event only when a handler was set.

This enables mlx5_ib to not set a handler upon CQ creation and use some
other mechanism to get this event as of other events by the
mlx5_eq_notifier_register API.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>

Enhance mlx5_core_create_cq() to get the command out buffer from the
callers to let them use the output.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>

Expose the API to register for ANY event, mlx5_ib will be able to use
this functionality for its needs.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>

Use the reported device capabilities for the supported user events (i.e.
affiliated and un-affiliated) to set the EQ mask.

As the event mask can be up to 256 defined by 4 entries of u64 change
the applicable code to work accordingly.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>

The firmware command to destroy a CQ might fail when the object is
referenced by other object and the ref count is managed by the firmware.

To enable a second successful destruction post the first failure need to
change  mlx5_eq_del_cq() to be a void function.

As an error in mlx5_eq_del_cq() is quite fatal from the option to
recover, a debug message inside it should be good enougth and it was
changed to be void.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>

Fixes gcc '-Wunused-but-set-variable' warning:

drivers/infiniband/hw/hns/hns_roce_hw_v2.c: In function 'hns_roce_function_clear':
drivers/infiniband/hw/hns/hns_roce_hw_v2.c:1135:7: warning:
 variable 'fclr_write_fail_flag' set but not used [-Wunused-but-set-variable]

It is never used, so can be removed.
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

The API for ib_query_qp requires the driver to set qp_state and
cur_qp_state on return, add the missing sets.

Fixes: d3749841 ("i40iw: add files for iwarp interface")
Signed-off-by: Changcheng Liu <changcheng.liu@aliyun.com>
Acked-by: Shiraz Saleem <shiraz.saleem@intel.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

Use kmemdump instead of kzmalloc + memcpy.
Signed-off-by: Fuqian Huang <huangfq.daxian@gmail.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

vzalloc has already zeroed the memory.  So a memset is unneeded.
Signed-off-by: Fuqian Huang <huangfq.daxian@gmail.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

In commit af7ddd8a ("Merge tag 'dma-mapping-4.21' of
git://git.infradead.org/users/hch/dma-mapping"),
dma_alloc_coherent/dmam_alloc_coherent always zeroed the returned memory.
So the memset after a coherent allocation function is not needed.
Signed-off-by: Fuqian Huang <huangfq.daxian@gmail.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

Bernard Metzler says:

====================
This patch set contributes the SoftiWarp driver rebased for latest
rdma-next. SoftiWarp (siw) implements the iWarp RDMA protocol over kernel
TCP sockets. The driver integrates with the linux-rdma framework.

A matching userlevel driver is available as PR at
https://github.com/linux-rdma/rdma-core/pull/536

Many thanks for reviewing and testing the driver, especially to Leon,
Jason, Steve, Doug, Olga, Dennis, Gal. You all helped to significantly
improve the driver over the last year.

Please find below a list of changes and comments, compared to older
versions of the siw driver.

Many thanks!
Bernard.

CHANGES:
========

v3 (this version)
-----------------

- Rebased to rdma-next

- Removed unneccessary initialization of enums in siw-abi.h

- Added comment on sizing of all work queues to power of two.

v2
-----------------

- Changed recieve path CRC calculation to compute CRC32c not
  on target buffer after placement, but on original skbuf.
  This change severely hurts performance, if CRC is switched
  on, since skb must now be walked twice. It is planned to
  work on an extension to skb_copy_bits() to fold in CRC
  computation.

- Moved debugging to using ibdev_dbg().

- Dropped detailed packet debug printing.

- Removed siw_debug.[ch] files.

- Removed resource tracking, code now relies on restrack of
  RDMA midlayer. Only object counting to enforce reported
  device limits is left in place.

- Removed all nested switch-case statements.

- Cleaned up header file #include's

- Moved CQ create/destroy to new semantics,
  where midlayer creates/destroys containing object.

- Set siw's ABI version to 1 (was 0 before)

- Removed all enum initialization where not needed.

- Fixed MAINTANERS entry for siw driver

- This version stays with the current siw specific
  management of user memory (siw_umem_get() vs.
  ib_umem_get(), etc.). This, since the current ib_umem
  implementation is less efficient for user page lookup
  on the fast path, where effciency is important for a
  SW RDMA driver.
  It is planned to contribute enhancements to the ib_umem
  framework, wich makes it suitable for SW drivers as well.

v1 (first version after v9 of siw RFC)
--------------------------------------

- Rebased to 5.2-rc1

- All IDR code got removed.

- Both MR and QP deallocation verbs now synchronously
  free the resources referenced by the RDMA mid-layer.

- IPv6 support was added.

- For compatibility with Chelsio iWarp hardware, the RX
  path was slightly reworked. It now allows packet intersection
  between tagged and untagged RDMAP operations. While not
  a defined behavior as of IETF RFC 5040/5041, some RDMA hardware
  may intersect an ongoing outbound (large) tagged message, such
  as an multisegment RDMA Read Response with sending an untagged
  message, such as an RDMA Send frame. This behavior was only
  detected in an NVMeF setup, where siw was used at target side,
  and RDMA hardware at client side (during file write). siw now
  implements two input paths for tagged and untagged messages each,
  and allows the intersected placement of both messages.

- The siw kernel abi file got renamed from siw_user.h to siw-abi.h.
====================

* branch 'siw':
  SIW addition to kernel build environment
  SIW completion queue methods
  SIW receive path
  SIW transmit path
  SIW queue pair methods
  SIW application buffer management
  SIW application interface
  SIW connection management
  SIW network and RDMA core interface
  SIW main include file
  iWarp wire packet format

Broken up commit to add the Soft iWarp RDMA driver.
Signed-off-by: Bernard Metzler <bmt@zurich.ibm.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>