linux-stable.git/include/linux/mlx5, branch linux-5.1.y

net/mlx5e: Rx, Fix checksum calculation for new hardware

2019-07-28T06:28:35+00:00

[ Upstream commit db849faa9bef993a1379dc510623f750a72fa7ce ]

CQE checksum full mode in new HW, provides a full checksum of rx frame.
Covering bytes starting from eth protocol up to last byte in the received
frame (frame_size - ETH_HLEN), as expected by the stack.

Fixing up skb->csum by the driver is not required in such case. This fix
is to avoid wrong checksum calculation in drivers which already support
the new hardware with the new checksum mode.

Fixes: 85327a9c4150 ("net/mlx5: Update the list of the PCI supported devices")
Signed-off-by: Saeed Mahameed 
Signed-off-by: Greg Kroah-Hartman

net/mlx5: E-Switch, Use atomic rep state to serialize state change

2019-05-31T13:43:17+00:00

[ Upstream commit 6f4e02193c9a9ea54dd3151cf97489fa787cd0e6 ]

When the state of rep was introduced, it was also designed to prevent
duplicate unloading of the same rep. Considering the following two
flows when an eswitch manager is at switchdev mode with n VF reps loaded.

+--------------------------------------+--------------------------------+
| cpu-0                                | cpu-1                          |
| --------                             | --------                       |
| mlx5_ib_remove                       | mlx5_eswitch_disable_sriov     |
|  mlx5_ib_unregister_vport_reps       |  esw_offloads_cleanup          |
|   mlx5_eswitch_unregister_vport_reps |   esw_offloads_unload_all_reps |
|    __unload_reps_all_vport           |    __unload_reps_all_vport     |
+--------------------------------------+--------------------------------+

These two flows will try to unload the same rep. Per original design,
once one flow unloads the rep, the state moves to REGISTERED. The 2nd
flow will no longer needs to do the unload and bails out. However, as
read and write of the state is not atomic, when 1st flow is doing the
unload, the state is still LOADED, 2nd flow is able to do the same
unload action. Kernel crash will happen.

To solve this, driver should do atomic test-and-set for the state. So
that only one flow can change the rep state from LOADED to REGISTERED,
and proceed to do the actual unloading.

Since the state is changing to atomic type, all other read/write should
be atomic action as well.

Fixes: f121e0ea9586 (net/mlx5: E-Switch, Add state to eswitch vport representors)
Signed-off-by: Bodong Wang 
Reviewed-by: Parav Pandit 
Reviewed-by: Vu Pham 
Signed-off-by: Saeed Mahameed 
Signed-off-by: Sasha Levin

RDMA/mlx5: Use get_zeroed_page() for clock_info

2019-05-25T16:16:30+00:00

commit ddcdc368b1033e19fd3a5f750752e10e28a87826 upstream.

get_zeroed_page() returns a virtual address for the page which is better
than allocating a struct page and doing a permanent kmap on it.

Cc: stable@vger.kernel.org
Signed-off-by: Jason Gunthorpe 
Reviewed-by: Haggai Eran 
Signed-off-by: Leon Romanovsky 
Signed-off-by: Jason Gunthorpe 
Signed-off-by: Greg Kroah-Hartman

net/mlx5e: Add a lock on tir list

2019-03-29T19:24:41+00:00

Refresh tirs is looping over a global list of tirs while netdevs are
adding and removing tirs from that list. That is why a lock is
required.

Fixes: 724b2aa15126 ("net/mlx5e: TIRs management refactoring")
Signed-off-by: Yuval Avnery 
Signed-off-by: Saeed Mahameed

IB/mlx5: Use mlx5 core to create/destroy a DEVX DCT

2019-03-18T00:40:39+00:00

To prevent a hardware memory leak when a DEVX DCT object is destroyed
without calling DRAIN DCT before, (e.g. under cleanup flow), need to
manage its creation and destruction via mlx5 core.

In that case the DRAIN DCT command will be called and only once that it
will be completed the DESTROY DCT command will be called.  Otherwise, the
DESTROY DCT may fail and a hardware leak may occur.

As of that change the DRAIN DCT command should not be exposed any more
from DEVX, it's managed internally by the driver to work as expected by
the device specification.

Fixes: 7efce3691d33 ("IB/mlx5: Add obj create and destroy functionality")
Signed-off-by: Yishai Hadas 
Reviewed-by: Artemy Kovalyov 
Signed-off-by: Leon Romanovsky 
Signed-off-by: Jason Gunthorpe

Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma

2019-03-09T23:53:03+00:00

Pull rdma updates from Jason Gunthorpe:
 "This has been a slightly more active cycle than normal with ongoing
  core changes and quite a lot of collected driver updates.

   - Various driver fixes for bnxt_re, cxgb4, hns, mlx5, pvrdma, rxe

   - A new data transfer mode for HFI1 giving higher performance

   - Significant functional and bug fix update to the mlx5
     On-Demand-Paging MR feature

   - A chip hang reset recovery system for hns

   - Change mm->pinned_vm to an atomic64

   - Update bnxt_re to support a new 57500 chip

   - A sane netlink 'rdma link add' method for creating rxe devices and
     fixing the various unregistration race conditions in rxe's
     unregister flow

   - Allow lookup up objects by an ID over netlink

   - Various reworking of the core to driver interface:
       - drivers should not assume umem SGLs are in PAGE_SIZE chunks
       - ucontext is accessed via udata not other means
       - start to make the core code responsible for object memory
         allocation
       - drivers should convert struct device to struct ib_device via a
         helper
       - drivers have more tools to avoid use after unregister problems"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma: (280 commits)
  net/mlx5: ODP support for XRC transport is not enabled by default in FW
  IB/hfi1: Close race condition on user context disable and close
  RDMA/umem: Revert broken 'off by one' fix
  RDMA/umem: minor bug fix in error handling path
  RDMA/hns: Use GFP_ATOMIC in hns_roce_v2_modify_qp
  cxgb4: kfree mhp after the debug print
  IB/rdmavt: Fix concurrency panics in QP post_send and modify to error
  IB/rdmavt: Fix loopback send with invalidate ordering
  IB/iser: Fix dma_nents type definition
  IB/mlx5: Set correct write permissions for implicit ODP MR
  bnxt_re: Clean cq for kernel consumers only
  RDMA/uverbs: Don't do double free of allocated PD
  RDMA: Handle ucontext allocations by IB/core
  RDMA/core: Fix a WARN() message
  bnxt_re: fix the regression due to changes in alloc_pbl
  IB/mlx4: Increase the timeout for CM cache
  IB/core: Abort page fault handler silently during owning process exit
  IB/mlx5: Validate correct PD before prefetch MR
  IB/mlx5: Protect against prefetch of invalid MR
  RDMA/uverbs: Store PR pointer before it is overwritten
  ...

net/mlx5: Emit port affinity event for multipath offloads

2019-03-01T20:04:17+00:00

Under multipath offload scheme, as part of handling fib events, emit
mlx5 port affinity event on the enabled ports which will be handled by
the tc offloads code.

Signed-off-by: Roi Dayan 
Reviewed-by: Or Gerlitz 
Signed-off-by: Saeed Mahameed

net/mlx5: Add multipath mode

2019-03-01T20:04:16+00:00

In order to offload ecmp-on-host scheme where next-hop routes are used,
we will make use of HW LAG. Add accessor function to let upper layers
in the driver to realize if the lag acts in multi-path mode.

Signed-off-by: Roi Dayan 
Reviewed-by: Or Gerlitz 
Signed-off-by: Saeed Mahameed

net/mlx5e: Fix GRE key by controlling port tunnel entropy calculation

2019-02-22T21:38:23+00:00

Flow entropy is calculated on the inner packet headers and used for
flow distribution in processing, routing etc. For GRE-type
encapsulations the entropy value is placed in the eight LSB of the key
field in the GRE header as defined in NVGRE RFC 7637. For UDP based
encapsulations the entropy value is placed in the source port of the
UDP header.
The hardware may support entropy calculation specifically for GRE and
for all tunneling protocols. With commit df2ef3bff193 ("net/mlx5e: Add
GRE protocol offloading") GRE is offloaded, but the hardware is
configured by default to calculate flow entropy so packets transmitted
on the wire have a wrong key. To support UDP based tunnels (i.e VXLAN),
GRE (i.e. no flow entropy) and NVGRE (i.e. with flow entropy) the
hardware behaviour must be controlled by the driver.

Ensure port entropy calculation is enabled for offloaded VXLAN tunnels
and disable port entropy calculation in the presence of offloaded GRE
tunnels by monitoring the presence of entropy enabling tunnels (i.e
VXLAN) and entropy disabing tunnels (i.e GRE).

Fixes: df2ef3bff193 ("net/mlx5e: Add GRE protocol offloading")
Signed-off-by: Eli Britstein 
Reviewed-by: Oz Shlomo 
Reviewed-by: Roi Dayan 
Signed-off-by: Saeed Mahameed

net/mlx5: Introduce tunnel entropy control in PCMR register

2019-02-22T21:38:23+00:00

When using the device packet encapsulation offload, the device
calculates an entropy value, representing the inner packet headers. The
entropy field is placed inside the outer packet headers. For UDP-type
encapsulations, the entropy is placed in the source port field of the
UDP header. For GRE-type encapsulations, the entropy is placed in the 8
LSB of the key field in the GRE header. If the device does not recognize
the encapsulation type, the entropy is not placed in the packet.

Entropy setting can be controlled using PCMR register. if encapsulation
offload is not used force_entropy_cap should be set to 0x0. Entropy
setting is enabled/disabled using entropy_calc, and could be
additionally enabled/disabled for GRE encapsulation by entropy_gre_calc.

As a pre-step to automatically control the tunnel entropy, introduce
the entropy fields in the PCMR register with no functional change.

Signed-off-by: Eli Britstein 
Reviewed-by: Oz Shlomo 
Reviewed-by: Roi Dayan 
Signed-off-by: Saeed Mahameed