Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NDoug Ledford <dledford@redhat.com>

Applications must not assume that max_sge and max_sge_rd are the same,
Hence expose max_sge_rd correctly as well.
Reported-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NDoug Ledford <dledford@redhat.com>

There is little chance our memory allocation will fail, so we can
combine initializing the work structs with allocating them instead of
looping through all of them once to allocate and again to initialize.
Then when we need to actually find out if our device is up or in the
process of going down, have all of our work structs batched up, take the
spin_lock once and only once, and do all of the batch under the one
spin_lock invocation instead of incurring all of the locked memory cycles
we would otherwise incur to take/release the spin_lock over and over
again.
Signed-off-by: NDoug Ledford <dledford@redhat.com>

We create a number of work structs to be queued up to a workqueue, and
on completion of the workqueue handler, the workqueue handler frees the
allocated memory.  If, however, we don't queue the work struct because
the device is going down, then we need to free the memory ourselves.
Signed-off-by: NDoug Ledford <dledford@redhat.com>

On failure, we loop through all possible pointers and test them before
calling kfree.  But really, why even attempt to free items we didn't
allocate when we can easily loop through exactly and only the devices
for which the original memory allocation succeeded and free just those.
Signed-off-by: NManinder Singh <maninder1.s@samsung.com>
Signed-off-by: NDoug Ledford <dledford@redhat.com>

mlx4 VFs can provide CQE raw time-stamping services, but they
don't have the hca core clock mapped to their PCI bars.

As such, we should not attempt to query and report the clock offset
to user space for VFs. Doing so causes query_device over VFs to fail
with -ENOSUPP.

Fixes: 4b664c43 ('IB/mlx4: Add support for CQ time-stamping')
Signed-off-by: NMatan Barak <matanb@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDoug Ledford <dledford@redhat.com>

This is an infrastructure step to attach all the QPs opened from the
IB driver to a counter in order to collect VF stats from the PF using
those counters.

If the port's type is Ethernet, the counter policy demands two counters
per port (one for RoCE and one for Ethernet). The port default counter
(allocated in mlx4_core) is used for the Ethernet netdev QPs and we
allocate another counter for RoCE.

If the port's traffic is Infiniband, the counter policy demands
one counter per port, so it can use the port's default counter.

Also, Add 'allocated' flag for each counter in order to clean it at
unload.
Signed-off-by: NEran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: NHadar Hen Zion <hadarh@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add max MAD size to the device immutable data set and have all drivers that
support MADs report the current IB MAD size (IB_MGMT_MAD_SIZE) to the core.

Verify MAD size data in both the MAD core and when reading the immutable data.

OPA drivers will report alternate MAD sizes in subsequent patches.
Signed-off-by: NIra Weiny <ira.weiny@intel.com>
Signed-off-by: NDoug Ledford <dledford@redhat.com>

This includes:

* support allocation of CQ with the TIMESTAMP_COMPLETION creation flag.

* add timestamp_mask and hca_core_clock to query_device, reporting the
  number of supported timestamp bits (mask) and the hca_core_clock frequency.

* return hca core clock's offset in query_device vendor's data,
  this is needed in order to read the HCA's core clock.
Signed-off-by: NMatan Barak <matanb@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDoug Ledford <dledford@redhat.com>

In order to read the HCA's cycle counter efficiently in
user space, we need to map the HCA's register.
This is done through mmap call.
Signed-off-by: NMatan Barak <matanb@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDoug Ledford <dledford@redhat.com>

Vendors should be able to pass vendor specific data to/from
user-space via query_device uverb. In order to do this,
we need to pass the vendors' specific udata.
Signed-off-by: NMatan Barak <matanb@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDoug Ledford <dledford@redhat.com>

Currently, ib_create_cq uses cqe and comp_vecotr instead
of the extendible ib_cq_init_attr struct.

Earlier patches already changed the vendors to work with
ib_cq_init_attr. This patch changes the consumers too.
Signed-off-by: NMatan Barak <matanb@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDoug Ledford <dledford@redhat.com>

The unwinding clean up code are err_create_flow starts at the current
index i.  That means we shouldn't increment i until we're really sure
we won't have to destroy the current flow; otherwise we might
increment the index, fail inside an is_bonded block, and end up
accessing off the end of the reg_id[] array.

This was detected by Coverity (CID 1271229).
Signed-off-by: NRoland Dreier <roland@purestorage.com>
Signed-off-by: NDoug Ledford <dledford@redhat.com>

Previously, mlx4_en allocated EQs and used them exclusively.
This affected RoCE performance, as applications which are
events sensitive were limited to use only the legacy EQs.

Change that by introducing an EQ pool. This pool is managed
by mlx4_core. EQs are assigned to ports (when there are limited
number of EQs, multiple ports could be assigned to the same EQs).

An exception to this rule is the ASYNC EQ which handles various events.

Legacy EQs are completely removed as all EQs could be shared.

When a consumer (mlx4_ib/mlx4_en) requests an EQ, it asks for
EQ serving on a specific port. The core driver calculates which
EQ should be assigned to that request.

Because IRQs are shared between IB and Ethernet modules, their
names only include the PCI device BDF address.
Signed-off-by: NMatan Barak <matanb@mellanox.com>
Signed-off-by: NIdo Shamay <idos@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In SRIOV, when simple (i.e - Ethernet L2 only) flow steering rules are
created, always create them at MLX4_DOMAIN_NIC priority (instead of
the real priority the function created them at). This is done in order
to let multiple functions add broadcast/multicast rules without
affecting other functions, which is necessary for DPDK in SRIOV.
Signed-off-by: NMatan Barak <matanb@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Remove query_protocol callback

Use the new Core Capability bits for:

rdma_protocol_*
rdma_cap_ib_mad
rdma_cap_ib_smi
rdma_cap_ib_cm
rdma_cap_iw_cm
rdma_cap_ib_sa
rdma_cap_ib_mcast
rdma_cap_af_ib
rdma_cap_eth_ah
Signed-off-by: NIra Weiny <ira.weiny@intel.com>
Signed-off-by: NDoug Ledford <dledford@redhat.com>

As of commit 5eb620c8 "IB/core: Add helpers for uncached GID and P_Key
searches"; pkey_tbl_len and gid_tbl_len are immutable data which are stored in
the ib_device.

The per port core capability flags to be added later are also immutable data to
be stored in the ib_device object.

In preparation for this create a structure for per port immutable data and
place the pkey and gid table lengths within this structure.

"get_port_immutable" is added as a mandatory device function to allow the
drivers to fill in this data.
Signed-off-by: NIra Weiny <ira.weiny@intel.com>
Signed-off-by: NDoug Ledford <dledford@redhat.com>

Add new callback query_protocol() and implement for each HW.

Mapping List:
		node-type	link-layer	transport	protocol
nes		RNIC		ETH		IWARP		IWARP
amso1100	RNIC		ETH		IWARP		IWARP
cxgb3   	RNIC		ETH		IWARP		IWARP
cxgb4   	RNIC		ETH		IWARP		IWARP
usnic   	USNIC_UDP	ETH		USNIC_UDP	USNIC_UDP
ocrdma  	IB_CA		ETH		IB		IBOE
mlx4    	IB_CA		IB/ETH		IB		IB/IBOE
mlx5    	IB_CA		IB		IB		IB
ehca    	IB_CA		IB		IB		IB
ipath   	IB_CA		IB		IB		IB
mthca   	IB_CA		IB		IB		IB
qib     	IB_CA		IB		IB		IB
Signed-off-by: NMichael Wang <yun.wang@profitbricks.com>
Reviewed-by: NIra Weiny <ira.weiny@intel.com>
Tested-by: NIra Weiny <ira.weiny@intel.com>
Reviewed-by: NSean Hefty <sean.hefty@intel.com>
Reviewed-by: NJason Gunthorpe <jgunthorpe@obsidianresearch.com>
Tested-by: NDoug Ledford <dledford@redhat.com>
Signed-off-by: NDoug Ledford <dledford@redhat.com>

These KERN_<LEVEL> uses are unnecessary with pr_<level> and cause
bad logging output so remove them.
Signed-off-by: NJoe Perches <joe@perches.com>
Acked-by: NSteve Wise <swise@opengridcomputing.com>
Signed-off-by: NDoug Ledford <dledford@redhat.com>

Change the default mode to be HOST assigned instead of SM assigned. This is
the expected operational mode, because it doesn't depend on SM availability.

As PF generates random GUIDs as the initial admin values, this gives
out of the box experience.
Signed-off-by: NYishai Hadas <yishaih@mellanox.com>
Signed-off-by: NJack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDoug Ledford <dledford@redhat.com>

Request GIDs from the SM on demand, i.e., when a VF actually needs them,
and release them when the GIDs are no longer in use.

In cloud environments, this is useful for GID migrations, in which a
GID is assigned to a VF on the destination HCA, while the VF on the
source HCA is shutdown (but the GID was not administratively released).
Signed-off-by: NYishai Hadas <yishaih@mellanox.com>
Signed-off-by: NJack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDoug Ledford <dledford@redhat.com>

The calls to SET_PORT used hard-code numbers, when supplying command's
opcode modifiers, fix that to use well defined constants.
Signed-off-by: NIdo Shamay <idos@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Processing an event is done in a different context from the one when
the event was dispatched. This requires a check that the slave
net device is still valid when the event is being processed. The check is done
under the iboe lock which ensure correctness.

Fixes: a5750090 ('IB/mlx4: Add port aggregation support')
Signed-off-by: NMoni Shoua <monis@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The MLX4_PROT_IB_IPV4 protocol should only be used with RoCEv2 and such.
Removing this wrong usage allows to run multicast applications over RoCE.

Fixes: d487ee77 ("IB/mlx4: Use IBoE (RoCE) IP based GIDs in the port GID table")
Reported-by: NCarol Soto <clsoto@linux.vnet.ibm.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

The driver exposes interfaces that directly relate to HW state. Upon fatal
error, consumers of these interfaces (ULPs) that rely on completion of
all their posted work-request could hang, thereby introducing dependencies
in shutdown order.  To prevent this from happening, we manage the
relevant resources (CQs, QPs) that are used by the device. Upon a fatal error,
we now generate simulated completions for outstanding WQEs that were not
completed at the time the HW was reset.

It includes invoking the completion event handler for all involved CQs so that
the ULPs will poll those CQs. When polled we return simulated CQEs with
IB_WC_WR_FLUSH_ERR return code enabling ULPs to clean up their resources and
not wait forever for completions upon receiving remove_one.

The above change requires an extra check in the data path to make sure that when
device is in error state, the simulated CQEs will be returned and no further
WQEs will be posted.
Signed-off-by: NYishai Hadas <yishaih@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When attaching a QP to a multicast address in bonded mode, there was an
assumption that the port of the QP must be #1. This assumption isn't the
case under the flow which enables maximal usage of the physical ports.

Fix it by always checking the port of the original flow and create the
mirrored flow on the other port.

Fixes: c6215745 ('IB/mlx4: Load balance ports in port aggregation mode')
Signed-off-by: NMoni Shoua <monis@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When the mlx4 IB (RoCE) device works in link aggregation mode, it
exposes a single port to upper layers. Therefore, applications always
set '1' in port_num attribute when modifying a QP or creating an address handle.

To make sure that a node uses all available ports the mlx4 driver will
override the port_num attribute with a round robin policy.
Signed-off-by: NMoni Shoua <monis@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In port aggregation mode flows for port #1 (the only port) should be mirrored
on port #2. This is because packets can arrive from either physical ports.
Signed-off-by: NMoni Shoua <monis@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Register the interface with the mlx4 core driver with port aggregation support
and check for port aggregation mode when the 'add' function is called.

In this mode, only one physical port is reported to the upper layer
(RoCE/IB core stack and ULPs).
Signed-off-by: NMoni Shoua <monis@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Maintain a persistent memory that should survive reset flow/PCI error.
This comes as a preparation for coming series to support above flows.
Signed-off-by: NYishai Hadas <yishaih@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Except for VXLAN steering rules, all offloads should work as they were
under plain DMFS mode. Fix that by enabling all the offloads under
DMFS-A0 mode, except for VXLAN steering rules.

Fixes: d57febe1 "net/mlx4: Add A0 hybrid steering"
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When using BF (Blue-Flame), the QPN overrides the VLAN, CV, and SV fields
in the WQE. Thus, BF may only be used for QPNs with bits 6,7 unset.

The current Ethernet driver code reserves a Tx QP range with 256b alignment.

This is wrong because if there are more than 64 Tx QPs in use,
QPNs >= base + 65 will have bits 6/7 set.

This problem is not specific for the Ethernet driver, any entity that
tries to reserve more than 64 BF-enabled QPs should fail. Also, using
ranges is not necessary here and is wasteful.

The new mechanism introduced here will support reservation for
"Eth QPs eligible for BF" for all drivers: bare-metal, multi-PF, and VFs
(when hypervisors support WC in VMs). The flow we use is:

1. In mlx4_en, allocate Tx QPs one by one instead of a range allocation,
   and request "BF enabled QPs" if BF is supported for the function

2. In the ALLOC_RES FW command, change param1 to:
a. param1[23:0]  - number of QPs
b. param1[31-24] - flags controlling QPs reservation

Bit 31 refers to Eth blueflame supported QPs. Those QPs must have
bits 6 and 7 unset in order to be used in Ethernet.

Bits 24-30 of the flags are currently reserved.

When a function tries to allocate a QP, it states the required attributes
for this QP. Those attributes are considered "best-effort". If an attribute,
such as Ethernet BF enabled QP, is a must-have attribute, the function has
to check that attribute is supported before trying to do the allocation.

In a lower layer of the code, mlx4_qp_reserve_range masks out the bits
which are unsupported. If SRIOV is used, the PF validates those attributes
and masks out unsupported attributes as well. In order to notify VFs which
attributes are supported, the VF uses QUERY_FUNC_CAP command. This command's
mailbox is filled by the PF, which notifies which QP allocation attributes
it supports.
Signed-off-by: NEugenia Emantayev <eugenia@mellanox.co.il>
Signed-off-by: NMatan Barak <matanb@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Previously, the driver queried the firmware in order to get the number
of supported EQs. Under SRIOV, since this was done before the driver
notified the firmware how many VFs it actually needs, the firmware had
to take into account a worst case scenario and always allocated four EQs
per VF, where one was used for events while the others were used for completions.

Now, when the firmware supports the asymmetric allocation scheme, denoted
by exposing num_sys_eqs > 0 (--> MLX4_DEV_CAP_FLAG2_SYS_EQS), we use the
QUERY_FUNC command to query the firmware before enabling SRIOV. Thus we
can get more EQs and MSI-X vectors per function.

Moreover, when running in the new firmware/driver mode, the limitation
that the number of EQs should be a power of two is lifted.
Signed-off-by: NJack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: NMatan Barak <matanb@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

If mlx4_ib_create_flow() attempts to create > 1 rules with the
firmware, and one of these registrations fail, we leaked the
already created flow rules.

One example of the leak is when the registration of the VXLAN ghost
steering rule fails, we didn't unregister the original rule requested
by the user, introduced in commit d2fce8a9 "mlx4: Set
user-space raw Ethernet QPs to properly handle VXLAN traffic".

While here, add dump of the VXLAN portion of steering rules
so it can actually be seen when flow creation fails.
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We had several problems here.  First, a race condition on QP1 mac
handling between mlx4_ib_update_qps and mlx4_ib_modify_qp, which is
fixed by taking the qp mutex in mlx4_ib_update_qps.

Also, qp->pri.smac_port was not updated in mlx4_ib_update_qps.

Last, in __mlx4_ib_modify_qp we did not properly handle the case where
the mac is zero, but port is non-zero.
Signed-off-by: NJack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

Automatic Path Migration is not supported under RoCE. Therefore,
return a "not-supported" error if the caller attempts to set an
alternate path in a QP context.

In addition, if there are no IB ports configured, do not report
APM capability in the device flags returned by mlx4_ib_query_device.
Signed-off-by: NJack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

For native functions (non-SR-IOV), there's no reason to update
the smac_index, as QP1 is a GSI QP.
Signed-off-by: NJack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

The source MAC is needed in RoCE when building the QP1 header.

Currently, this is obtained from the source net device. However, the net
device may not yet exist, or can be destroyed in parallel to this QP1 send
operation (e.g through the VPI port change flow) so accessing it may cause
a kernel crash.

To fix this, we maintain a source MAC cache per port for the net device in
struct mlx4_ib_roce.  This cached MAC is initialized to be the default MAC
address obtained during HCA initialization via QUERY_PORT. This cached MAC
is updated via the netdev event notifier handler.

Since the cached MAC is held in an atomic64 object, we do not need locking
when accessing it.
Signed-off-by: NJack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

When device is being removed (e.g during VPI port link type change
from ETH to IB), tasks for gid table changes should not be executed.

Flush the current queue of tasks and block further tasks from entering the queue.
Signed-off-by: NMoni Shoua <monis@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

Chuck Lever reported the following stack trace:

    =================================
    [ INFO: inconsistent lock state ]
    3.16.0-rc2-00024-g2e78883 #17 Tainted: G            E
    ---------------------------------
    inconsistent {SOFTIRQ-ON-W} -> {IN-SOFTIRQ-W} usage.
    swapper/0/0 [HC0[0]:SC1[1]:HE1:SE0] takes:
    (&(&iboe->lock)->rlock){+.?...}, at: [<ffffffffa065f68b>] mlx4_ib_addr_event+0xdb/0x1a0 [mlx4_ib]
    {SOFTIRQ-ON-W} state was registered at:
     [<ffffffff810b3110>] mark_irqflags+0x110/0x170
     [<ffffffff810b4806>] __lock_acquire+0x2c6/0x5b0
     [<ffffffff810b4bd9>] lock_acquire+0xe9/0x120
     [<ffffffff815f7f6e>] _raw_spin_lock+0x3e/0x80
     [<ffffffffa0661084>] mlx4_ib_scan_netdevs+0x34/0x260 [mlx4_ib]
     [<ffffffffa06612db>] mlx4_ib_netdev_event+0x2b/0x40 [mlx4_ib]
     [<ffffffff81522219>] register_netdevice_notifier+0x99/0x1e0
     [<ffffffffa06626e3>] mlx4_ib_add+0x743/0xbc0 [mlx4_ib]
     [<ffffffffa05ec168>] mlx4_add_device+0x48/0xa0 [mlx4_core]
     [<ffffffffa05ec2c3>] mlx4_register_interface+0x73/0xb0 [mlx4_core]
     [<ffffffffa05c505e>] cm_req_handler+0x13e/0x460 [ib_cm]
     [<ffffffff810002e2>] do_one_initcall+0x112/0x1c0
     [<ffffffff810e8264>] do_init_module+0x34/0x190
     [<ffffffff810ea62f>] load_module+0x5cf/0x740
     [<ffffffff810ea939>] SyS_init_module+0x99/0xd0
     [<ffffffff815f8fd2>] system_call_fastpath+0x16/0x1b
    irq event stamp: 336142
    hardirqs last  enabled at (336142): [<ffffffff810612f5>] __local_bh_enable_ip+0xb5/0xc0
    hardirqs last disabled at (336141): [<ffffffff81061296>] __local_bh_enable_ip+0x56/0xc0
    softirqs last  enabled at (336004): [<ffffffff8106123a>] _local_bh_enable+0x4a/0x50
    softirqs last disabled at (336005): [<ffffffff810617a4>] irq_exit+0x44/0xd0

    other info that might help us debug this:
    Possible unsafe locking scenario:

          CPU0
          ----
     lock(&(&iboe->lock)->rlock);
     <Interrupt>
       lock(&(&iboe->lock)->rlock);

    *** DEADLOCK ***

The above problem was caused by the spin lock being taken both in the process
context and in a soft-irq context (in a netdev notifier handler).

The required fix is to use spin_lock/unlock_bh() instead of spin_lock/unlock
on the iboe lock.
Reported-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NJack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>