openwrt/staging/blogic.git
7 years agoIB/mlx5: Use blue flame register allocator in mlx5_ib
Eli Cohen [Tue, 3 Jan 2017 21:55:25 +0000 (23:55 +0200)]
IB/mlx5: Use blue flame register allocator in mlx5_ib

Make use of the blue flame registers allocator at mlx5_ib. Since blue
flame was not really supported we remove all the code that is related to
blue flame and we let all consumers to use the same blue flame register.
Once blue flame is supported we will add the code. As part of this patch
we also move the definition of struct mlx5_bf to mlx5_ib.h as it is only
used by mlx5_ib.

Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
7 years agonet/mlx5: Add interface to get reference to a UAR
Eli Cohen [Tue, 3 Jan 2017 21:55:24 +0000 (23:55 +0200)]
net/mlx5: Add interface to get reference to a UAR

A reference to a UAR is required to generate CQ or EQ doorbells. Since
CQ or EQ doorbells can all be generated using the same UAR area without
any effect on performance, we are just getting a reference to any
available UAR, If one is not available we allocate it but we don't waste
the blue flame registers it can provide and we will use them for
subsequent allocations.
We get a reference to such UAR and put in mlx5_priv so any kernel
consumer can make use of it.

Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
7 years agonet/mlx5: Introduce blue flame register allocator
Eli Cohen [Tue, 3 Jan 2017 21:55:23 +0000 (23:55 +0200)]
net/mlx5: Introduce blue flame register allocator

Here is an implementation of an allocator that allocates blue flame
registers. A blue flame register is used for generating send doorbells.
A blue flame register can be used to generate either a regular doorbell
or a blue flame doorbell where the data to be sent is written to the
device's I/O memory hence saving the need to read the data from memory.
For blue flame kind of doorbells to succeed, the blue flame register
need to be mapped as write combining. The user can specify what kind of
send doorbells she wishes to use. If she requested write combining
mapping but that failed, the allocator will fall back to non write
combining mapping and will indicate that to the user.
Subsequent patches in this series will make use of this allocator.

Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
7 years agoIB/mlx5: Fix retrieval of index to first hi class bfreg
Eli Cohen [Tue, 3 Jan 2017 21:55:22 +0000 (23:55 +0200)]
IB/mlx5: Fix retrieval of index to first hi class bfreg

First the function retrieving the index of the first hi latency class
blue flame register. High latency class bfregs are located right above
medium latency class bfregs.

Fixes: c1be5232d21d ('IB/mlx5: Fix micro UAR allocator')
Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
7 years agomlx5: Fix naming convention with respect to UARs
Eli Cohen [Tue, 3 Jan 2017 21:55:21 +0000 (23:55 +0200)]
mlx5: Fix naming convention with respect to UARs

This establishes a solid naming conventions for UARs. A UAR (User Access
Region) can have size identical to a system page or can be fixed 4KB
depending on a value queried by firmware. Each UAR always has 4 blue
flame register which are used to post doorbell to send queue. In
addition, a UAR has section used for posting doorbells to CQs or EQs. In
this patch we change names to reflect this conventions.

Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
7 years agoIB/mlx5: Fix error handling order in create_kernel_qp
Eli Cohen [Tue, 3 Jan 2017 21:55:20 +0000 (23:55 +0200)]
IB/mlx5: Fix error handling order in create_kernel_qp

Make sure order of cleanup is exactly the opposite of initialization.

Fixes: 9603b61de1ee ('mlx5: Move pci device handling from mlx5_ib to mlx5_core')
Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
7 years agoIB/mlx5: Fix kernel to user leak prevention logic
Eli Cohen [Tue, 3 Jan 2017 21:55:19 +0000 (23:55 +0200)]
IB/mlx5: Fix kernel to user leak prevention logic

The logic was broken as it failed to update the response length for
architectures with PAGE_SIZE larger than 4kB. As a result further
extension of the ucontext response struct would fail.

Fixes: d69e3bcf7976 ('IB/mlx5: Mmap the HCA's core clock register to user-space')
Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
7 years agoMerge branch 'mlx5-odp'
David S. Miller [Mon, 2 Jan 2017 20:51:21 +0000 (15:51 -0500)]
Merge branch 'mlx5-odp'

Saeed Mahameed says:

====================
Mellanox mlx5 core and ODP updates 2017-01-01

The following eleven patches mainly come from Artemy Kovalyov
who expanded mlx5 on-demand-paging (ODP) support. In addition
there are three cleanup patches which don't change any functionality,
but are needed to align codebase prior accepting other patches.

Memory region (MR) in IB can be huge and ODP (on-demand paging)
technique allows to use unpinned memory, which can be consumed and
released on demand. This allows to applications do not pin down
the underlying physical pages of the address space, and save from them
need to track the validity of the mappings.

Rather, the HCA requests the latest translations from the OS when pages
are not present, and the OS invalidates translations which are no longer
valid due to either non-present pages or mapping changes.

In existing ODP implementation applications is needed to register
memory buffers for communication, though registered memory regions
need not have valid mappings at registration time.

This patch set performs the following steps to expand
current ODP implementation:

1. It refactors UMR to support large regions, by introducing generic
   function to perform HCA translation table modifications. This
   function supports both atomic and process contexts and is not limited
   by number of modified entries.

   This function allows to enable reallocated memory regions of
   arbitrary size, so adding MR cache buckets to support up to 16GB MRs.

2. It changes page fault event format and refactor page faults logic
   together with addition of atomic support.

3. It prepares mlx5 core code to support implicit registration with
   simplified and relaxed semantics.

   Implicit ODP semantics allows to applications provide special memory
   key that represents their complete address space. Thus all IO accesses
   referencing to this key (with proper access rights associated with the key)
   wouldn't need not register any virtual address range.

Thanks,
        Artemy, Ilya and Leon

v1->v2:
  - Don't use 'inline' in .c files
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoIB/mlx5: Improve MR check
Artemy Kovalyov [Mon, 2 Jan 2017 09:37:48 +0000 (11:37 +0200)]
IB/mlx5: Improve MR check

Add "type" field to mlx5_core MKEY struct.
Check whether page fault happens on MKEY corresponding to MR.

Signed-off-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoIB/mlx5: Add ODP atomics support
Artemy Kovalyov [Mon, 2 Jan 2017 09:37:47 +0000 (11:37 +0200)]
IB/mlx5: Add ODP atomics support

Handle ODP atomic operations. When initiator of RDMA atomic
operation use ODP MR to provide source data handle pagefault properly.

Signed-off-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years ago{net,IB}/mlx5: Refactor page fault handling
Artemy Kovalyov [Mon, 2 Jan 2017 09:37:46 +0000 (11:37 +0200)]
{net,IB}/mlx5: Refactor page fault handling

* Update page fault event according to last specification.
* Separate code path for page fault EQ, completion EQ and async EQ.
* Move page fault handling work queue from mlx5_ib static variable
  into mlx5_core page fault EQ.
* Allocate memory to store ODP event dynamically as the
  events arrive, since in atomic context - use mempool.
* Make mlx5_ib page fault handler run in process context.

Signed-off-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet/mlx5: Update PAGE_FAULT_RESUME layout
Artemy Kovalyov [Mon, 2 Jan 2017 09:37:45 +0000 (11:37 +0200)]
net/mlx5: Update PAGE_FAULT_RESUME layout

Update PAGE_FAULT_RESUME command layout.

Three bit fields describing page fault: rdma, rdma_write, req_res gave 8
possible combinations, while only a few were legal. Now they
are interpreted as three-bit type field, where former legal
combinations turns into corresponding types and unused were added as new
types.

Signed-off-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoIB/mlx5: Add MR cache for large UMR regions
Artemy Kovalyov [Mon, 2 Jan 2017 09:37:44 +0000 (11:37 +0200)]
IB/mlx5: Add MR cache for large UMR regions

In this change we turn mlx5_ib_update_mtt() into generic
mlx5_ib_update_xlt() to perfrom HCA translation table modifiactions
supporting both atomic and process contexts and not limited by number
of modified entries.
Using this function we increase preallocated MRs up to 16GB.

Signed-off-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoIB/mlx5: Add support for big MRs
Artemy Kovalyov [Mon, 2 Jan 2017 09:37:43 +0000 (11:37 +0200)]
IB/mlx5: Add support for big MRs

Make use of extended UMR translation offset.

Signed-off-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoIB/mlx5: Refactor UMR post send format
Artemy Kovalyov [Mon, 2 Jan 2017 09:37:42 +0000 (11:37 +0200)]
IB/mlx5: Refactor UMR post send format

* Update struct mlx5_wqe_umr_ctrl_seg.
* Currenlty UMR send_flags aim only certain use cases: enabled/disable
  cached MR, modifying XLT for ODP. By making flags independent make UMR
  more flexible allowing arbitrary manipulations.
* Since different UMR formats have different entry sizes UMR request
  should receive exact size of translation table update instead of
  number of entries. Rename field npages to xlt_size in struct mlx5_umr_wr
  and update relevant code accordingly.
* Add support of length64 bit.

Signed-off-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet/mlx5: Support new MR features
Artemy Kovalyov [Mon, 2 Jan 2017 09:37:41 +0000 (11:37 +0200)]
net/mlx5: Support new MR features

This patch adds the following items to IFC file.

1. MLX5_MKC_ACCESS_MODE_KSM enum value for creating KSM memory keys.
KSM access mode used when indirect MKey associated with fixed memory
size entries.

2. null_mkey field that is used to indicate non-present KLM/KSM
entries, where it causes the device to generate page fault event
when trying to access it.

3. struct mlx5_ifc_cmd_hca_cap_bits capability bits indicating
related value/field is supported:
* fixed_buffer_size - MLX5_MKC_ACCESS_MODE_KSM
* umr_extended_translation_offset - translation_offset_42_16
    in UMR ctrl segment
* null_mkey - null_mkey in QUERY_SPECIAL_CONTEXTS

Signed-off-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoIB/mlx5: Add helper mlx5_ib_post_send_wait
Binoy Jayan [Mon, 2 Jan 2017 09:37:40 +0000 (11:37 +0200)]
IB/mlx5: Add helper mlx5_ib_post_send_wait

Clean up the following common code (to post a list of work requests to the
send queue of the specified QP) at various places and add a helper function
'mlx5_ib_post_send_wait' to implement the same.

 - Initialize 'mlx5_ib_umr_context' on stack
 - Assign "mlx5_umr_wr:wr:wr_cqe to umr_context.cqe
 - Acquire the semaphore
 - call ib_post_send with a single ib_send_wr
 - wait_for_completion()
 - Check for umr_context.status
 - Release the semaphore

Signed-off-by: Binoy Jayan <binoy.jayan@linaro.org>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoIB/mlx5: Reorder code in query device command
Leon Romanovsky [Mon, 2 Jan 2017 09:37:39 +0000 (11:37 +0200)]
IB/mlx5: Reorder code in query device command

The order of features exposed by private mlx5-abi.h
file is CQE zipping, packet pacing and multi-packet WQE.

The internal order implemented in mlx5_ib_query_device() is
multi-packet WQE, CQE zipping and packet pacing.

Such difference hurts code readability, so let's sync,
while mlx5-abi.h (exposed to userspace) is the primary
order.

This commit doesn't change any functionality.

Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet/mlx5: Fix offset naming for reserved fields in hca_cap_bits
Max Gurtovoy [Mon, 2 Jan 2017 09:37:38 +0000 (11:37 +0200)]
net/mlx5: Fix offset naming for reserved fields in hca_cap_bits

Fix offset for reserved fields.

Fixes: 7486216b3a0b ("{net,IB}/mlx5: mlx5_ifc updates")
Fixes: b4ff3a36d3e4 ("net/mlx5: Use offset based reserved field names in the IFC header file")
Fixes: 7d5e14237a55 ("net/mlx5: Update mlx5_ifc hardware features")
Signed-off-by: Max Gurtovoy <maxg@mellanox.com>
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge tag 'wireless-drivers-next-for-davem-2017-01-02' of git://git.kernel.org/pub...
David S. Miller [Mon, 2 Jan 2017 20:23:34 +0000 (15:23 -0500)]
Merge tag 'wireless-drivers-next-for-davem-2017-01-02' of git://git./linux/kernel/git/kvalo/wireless-drivers-next

Kalle Valo says:

====================
wireless-drivers-next patches for 4.11

The most notable change here is the inclusion of airtime fairness
scheduling to ath9k. It prevents slow clients from hogging all the
airtime and unfairly slowing down faster clients.

Otherwise smaller changes and cleanup.

Major changes:

ath9k

* cleanup eeprom endian handling
* add airtime fairness scheduling

ath10k

* fix issues for new QCA9377 firmware version
* support dev_coredump() for firmware crash dump
* enable channel 169 on 5 GHz band
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: stmmac: remove unused duplicate property snps,axi_all
Niklas Cassel [Fri, 30 Dec 2016 12:56:46 +0000 (13:56 +0100)]
net: stmmac: remove unused duplicate property snps,axi_all

For core revision 3.x Address-Aligned Beats is available in two registers.
The DT property snps,aal was created for AAL in the DMA bus register,
which is a read/write bit.
The DT property snps,axi_all was created for AXI_AAL in the AXI bus mode
register, which is a read only bit that reflects the value of AAL in the
DMA bus register.

Since the value of snps,axi_all is never used in the driver,
and since the property was created for a bit that is read only,
it should be safe to remove the property.

Acked-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: Niklas Cassel <niklas.cassel@axis.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge branch 'qed-driver-updates'
David S. Miller [Mon, 2 Jan 2017 02:02:21 +0000 (21:02 -0500)]
Merge branch 'qed-driver-updates'

Yuval Mintz says:

====================
qed*: Driver updates

The more interesting changes in this series include:
  - Restructuring of the qede files - qede_main.c has grown big and this
    series splits it into 3 parts [patches #2 and #3].
  - Some significant changes in the API through which RSS indirection
    table gets configured [#8].
  - Support for ndo_set_vf_trust() [#9] which would regulate which VFs
    are allowed to use promisc/multi-promisc mode.

It also contains various minor changes to qed/qede, as well as
non-functional changes [#1, #12] to complement other changes.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoqed*: Advance driver versions to 8.10.10.20.
Mintz, Yuval [Sun, 1 Jan 2017 11:57:11 +0000 (13:57 +0200)]
qed*: Advance driver versions to 8.10.10.20.

Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoqed: Conserve RDMA resources when !QEDR
Ram Amrani [Sun, 1 Jan 2017 11:57:10 +0000 (13:57 +0200)]
qed: Conserve RDMA resources when !QEDR

If qedr isn't part of the kernel then don't allocate RDMA resources
for it in qed.

Signed-off-by: Ram Amrani <Ram.Amrani@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoqed: Support Multicast on Tx-switching
Mintz, Yuval [Sun, 1 Jan 2017 11:57:09 +0000 (13:57 +0200)]
qed: Support Multicast on Tx-switching

Currently multicast traffic wouldn't be routed internally to
listener; Instead it would only be sent to network via the
physical carrier.

Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoqed*: Add support for ndo_set_vf_trust
Mintz, Yuval [Sun, 1 Jan 2017 11:57:08 +0000 (13:57 +0200)]
qed*: Add support for ndo_set_vf_trust

Trusted VFs would be allowed to receive promiscuous and
multicast promiscuous data.

Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoqed*: RSS indirection based on queue-handles
Mintz, Yuval [Sun, 1 Jan 2017 11:57:07 +0000 (13:57 +0200)]
qed*: RSS indirection based on queue-handles

A step toward having qede agnostic to the queue configurations
in firmware/hardware - let the RSS indirections use queue handles
instead of actual queue indices.

Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoqede: Remove unnecessary datapath dereference
Mintz, Yuval [Sun, 1 Jan 2017 11:57:06 +0000 (13:57 +0200)]
qede: Remove unnecessary datapath dereference

Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoqede - mark SKB as encapsulated
Manish Chopra [Sun, 1 Jan 2017 11:57:05 +0000 (13:57 +0200)]
qede - mark SKB as encapsulated

When driver receives a recognized encapsulated packet it needs
to set the skb->encapsulation field as well.

Signed-off-by: Manish Chopra <Manish.Chopra@cavium.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoqede: Postpone reallocation until NAPI end
Mintz, Yuval [Sun, 1 Jan 2017 11:57:04 +0000 (13:57 +0200)]
qede: Postpone reallocation until NAPI end

During Rx flow driver allocates a replacement buffer each time
it consumes an Rx buffer. Failing to do so, it would consume the
currently processed buffer and re-post it on the ring.
As a result, the Rx ring is always completely full [from driver POV].

We now allow the Rx ring to shorten by doing the re-allocations
at the end of the NAPI run. The only limitation is that we still want to
make sure each time we reallocate that we'd still have sufficient
elements in the Rx ring to guarantee that FW would be able to post
additional data and trigger an interrupt.

Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoqed*: Change maximal number of queues
Mintz, Yuval [Sun, 1 Jan 2017 11:57:03 +0000 (13:57 +0200)]
qed*: Change maximal number of queues

Today qede requests contexts that would suffice for 64 'whole'
combined queues [192 meant for 64 rx, tx and xdp tx queues],
but registers netdev and limits the number of queues based on
information received by qed. In turn, qed doesn't take context
into account when informing qede how many queues it can support.

This would lead to a configuration problem in case user tries
configuring >64 combined queues to interface [or >96 in case
xdp isn't enabled]. Since we don't have a mangement firware
that actually provides so many interrupt lines to a single
device we're currently safe but that's about to change soon.

The new maximum is hence changed:
  - For RoCE devices, the limit would remain 64.
  - For non-RoCE devices, the limit might be higher [depending
    on the actual configuration of the device].
qed would start enforcing that limit in both scenarios.

Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoqede: Split filtering logic to its own file
Mintz, Yuval [Sun, 1 Jan 2017 11:57:02 +0000 (13:57 +0200)]
qede: Split filtering logic to its own file

This takes the various filtering logic of the driver and
moves them into their own dedicated file - qede_filter.c.

Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoqede: Break datapath logic into its own file
Mintz, Yuval [Sun, 1 Jan 2017 11:57:01 +0000 (13:57 +0200)]
qede: Break datapath logic into its own file

This adds a new file qede_fp.c and relocates the datapath-related
logic into it [from qede_main.c].

Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoqed*: Update to dual-license
Mintz, Yuval [Sun, 1 Jan 2017 11:57:00 +0000 (13:57 +0200)]
qed*: Update to dual-license

Since the submission of the qedr driver, there's inconsistency
in the licensing of the various qed/qede files - some are GPLv2
and some are dual-license.
Since qedr requires dual-license and it's dependent on both,
we're updating the licensing of all qed/qede source files.

Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agortlwifi: fix spelling mistake: "encrypiton" -> "encryption"
Colin Ian King [Fri, 30 Dec 2016 14:50:27 +0000 (14:50 +0000)]
rtlwifi: fix spelling mistake: "encrypiton" -> "encryption"

trivial fix to spelling mistake in RT_TRACE message

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agowlcore: fix spelling mistake in wl1271_warning
Colin Ian King [Thu, 29 Dec 2016 20:14:00 +0000 (20:14 +0000)]
wlcore: fix spelling mistake in wl1271_warning

trivial fix to spelling mistake of function name in wl1271_warning,
should be dynamic_ps_timeout instead of dyanmic_ps_timeout.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agoMerge ath-next from git://git.kernel.org/pub/scm/linux/kernel/git/kvalo/ath.git
Kalle Valo [Sun, 1 Jan 2017 18:48:37 +0000 (20:48 +0200)]
Merge ath-next from git://git./linux/kernel/git/kvalo/ath.git

ath.git patches for 4.11. Major changes:

ath9k

* cleanup eeprom endian handling
* add airtime fairness scheduling

ath10k

* fix issues for new QCA9377 firmware version
* support dev_coredump() for firmware crash dump
* enable channel 169 on 5 GHz band

7 years agonet: 3com: typhoon: typhoon_init_one: make return values more specific
Thomas Preisner [Fri, 30 Dec 2016 02:37:54 +0000 (03:37 +0100)]
net: 3com: typhoon: typhoon_init_one: make return values more specific

In some cases the return value of a failing function is not being used
and the function typhoon_init_one() returns another negative error code
instead.

Signed-off-by: Thomas Preisner <thomas.preisner+linux@fau.de>
Signed-off-by: Milan Stephan <milan.stephan+linux@fau.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: 3com: typhoon: typhoon_init_one: fix incorrect return values
Thomas Preisner [Fri, 30 Dec 2016 02:37:53 +0000 (03:37 +0100)]
net: 3com: typhoon: typhoon_init_one: fix incorrect return values

In a few cases the err-variable is not set to a negative error code if a
function call in typhoon_init_one() fails and thus 0 is returned
instead.
It may be better to set err to the appropriate negative error
code before returning.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=188841

Reported-by: Pan Bian <bianpan2016@163.com>
Signed-off-by: Thomas Preisner <thomas.preisner+linux@fau.de>
Signed-off-by: Milan Stephan <milan.stephan+linux@fau.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: Allow IP_MULTICAST_IF to set index to L3 slave
David Ahern [Thu, 29 Dec 2016 23:39:37 +0000 (15:39 -0800)]
net: Allow IP_MULTICAST_IF to set index to L3 slave

IP_MULTICAST_IF fails if sk_bound_dev_if is already set and the new index
does not match it. e.g.,

    ntpd[15381]: setsockopt IP_MULTICAST_IF 192.168.1.23 fails: Invalid argument

Relax the check in setsockopt to allow setting mc_index to an L3 slave if
sk_bound_dev_if points to an L3 master.

Make a similar change for IPv6. In this case change the device lookup to
take the rcu_read_lock avoiding a refcnt. The rcu lock is also needed for
the lookup of a potential L3 master device.

This really only silences a setsockopt failure since uses of mc_index are
secondary to sk_bound_dev_if if it is set. In both cases, if either index
is an L3 slave or master, lookups are directed to the same FIB table so
relaxing the check at setsockopt time causes no harm.

Patch is based on a suggested change by Darwin for a problem noted in
their code base.

Suggested-by: Darwin Dingel <darwin.dingel@alliedtelesis.co.nz>
Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobrcmfmac: fix spelling mistakes on "Ivalid"
Colin Ian King [Fri, 23 Dec 2016 00:43:22 +0000 (00:43 +0000)]
brcmfmac: fix spelling mistakes on "Ivalid"

Trivial fixes to spelling mistake "Ivalid" to "Invalid" in
brcmf_err error messages.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agortlwifi: fix spelling mistake: "contry" -> "country"
Colin Ian King [Thu, 29 Dec 2016 16:00:29 +0000 (16:00 +0000)]
rtlwifi: fix spelling mistake: "contry" -> "country"

trivial fix to spelling mistake in RT_TRACE message

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agortlwifi: Fix alignment issues
Ping-Ke Shih [Wed, 28 Dec 2016 21:40:04 +0000 (15:40 -0600)]
rtlwifi: Fix alignment issues

The addresses of Wlan NIC registers are natural alignment, but some
drivers have bugs. These are evident on platforms that need natural
alignment to access registers.  This change contains the following:
 1. Function _rtl8821ae_dbi_read() is used to read one byte from DBI,
    thus it should use rtl_read_byte().
 2. Register 0x4C7 of 8192ee is single byte.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Stable <stable@vger.kernel.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agolibertas: constify cfg80211_ops structures
Bhumika Goyal [Sat, 17 Dec 2016 22:57:24 +0000 (04:27 +0530)]
libertas: constify cfg80211_ops structures

cfg80211_ops structures are only passed as an argument to the function
wiphy_new. This argument is of type const, so cfg80211_ops strutures
having this property can be declared as const.
Done using Coccinelle

@r1 disable optional_qualifier @
identifier i;
position p;
@@
static struct cfg80211_ops i@p = {...};

@ok1@
identifier r1.i;
position p;
@@
wiphy_new(&i@p,...)

@bad@
position p!={r1.p,ok1.p};
identifier r1.i;
@@
i@p

@depends on !bad disable optional_qualifier@
identifier r1.i;
@@
+const
struct cfg80211_ops i;

File size before:
   text    data     bss     dec     hex filename
  21225    1954      16   23195    5a9b wireless/marvell/libertas/cfg.o

File size after:
   text    data     bss     dec     hex filename
  22041    1154      16   23211    5aab wireless/marvell/libertas/cfg.o

Signed-off-by: Bhumika Goyal <bhumirks@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agortlwifi: Remove some redundant code
Larry Finger [Thu, 15 Dec 2016 18:23:10 +0000 (12:23 -0600)]
rtlwifi: Remove some redundant code

The symbol DBG_EMERG is no longer used and is removed.

In a number of places, the code has redundant messages. For example, if
the failure for the firmware to run is logged, it is not necessary to
log that the firmware has been started. In addition, extraneous braces are
removed.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agortlwifi: rtl8188ee: Remove all instances of DBG_EMERG
Larry Finger [Thu, 15 Dec 2016 18:23:09 +0000 (12:23 -0600)]
rtlwifi: rtl8188ee: Remove all instances of DBG_EMERG

This is a step toward eliminating the RT_TRACE macros. Those calls that
have DBG_EMERG as the level are always logged, and they represent error
conditions, thus they are replaced with pr_err().

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agortlwifi: rtl8192c-common: Remove all instances of DBG_EMERG
Larry Finger [Thu, 15 Dec 2016 18:23:08 +0000 (12:23 -0600)]
rtlwifi: rtl8192c-common: Remove all instances of DBG_EMERG

This is a step toward eliminating the RT_TRACE macros. Those calls that
have DBG_EMERG as the level are always logged, and they represent error
conditions, thus they are replaced with pr_err().

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agortlwifi: rtl8192ce: Remove all instances of DBG_EMERG
Larry Finger [Thu, 15 Dec 2016 18:23:07 +0000 (12:23 -0600)]
rtlwifi: rtl8192ce: Remove all instances of DBG_EMERG

This is a step toward eliminating the RT_TRACE macros. Those calls that
have DBG_EMERG as the level are always logged, and they represent error
conditions, thus they are replaced with pr_err().

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agortlwifi: rtl8192cu: Remove all instances of DBG_EMERG
Larry Finger [Thu, 15 Dec 2016 18:23:06 +0000 (12:23 -0600)]
rtlwifi: rtl8192cu: Remove all instances of DBG_EMERG

This is a step toward eliminating the RT_TRACE macros. Those calls that
have DBG_EMERG as the level are always logged, and they represent error
conditions, thus they are replaced with pr_err().

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agortlwifi: rtl8192de: Remove all instances of DBG_EMERG
Larry Finger [Thu, 15 Dec 2016 18:23:05 +0000 (12:23 -0600)]
rtlwifi: rtl8192de: Remove all instances of DBG_EMERG

This is a step toward eliminating the RT_TRACE macros. Those calls that
have DBG_EMERG as the level are always logged, and they represent error
conditions, thus they are replaced with pr_err().

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agortlwifi: rtl8192se: Remove all instances of DBG_EMERG
Larry Finger [Thu, 15 Dec 2016 18:23:04 +0000 (12:23 -0600)]
rtlwifi: rtl8192se: Remove all instances of DBG_EMERG

This is a step toward eliminating the RT_TRACE macros. Those calls that
have DBG_EMERG as the level are always logged, and they represent error
conditions, thus they are replaced with pr_err().

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agortlwifi: rtl8723-common: Remove all instances of DBG_EMERG
Larry Finger [Thu, 15 Dec 2016 18:23:03 +0000 (12:23 -0600)]
rtlwifi: rtl8723-common: Remove all instances of DBG_EMERG

This is a step toward eliminating the RT_TRACE macros. Those calls that
have DBG_EMERG as the level are always logged, and they represent error
conditions, thus they are replaced with pr_err().

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agortlwifi: rtl8192ee: Remove all instances of DBG_EMERG
Larry Finger [Thu, 15 Dec 2016 18:23:02 +0000 (12:23 -0600)]
rtlwifi: rtl8192ee: Remove all instances of DBG_EMERG

This is a step toward eliminating the RT_TRACE macros. Those calls that
have DBG_EMERG as the level are always logged, and they represent error
conditions, thus they are replaced with pr_err().

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agortlwifi: rtl8723ae: Remove all instances of DBG_EMERG
Larry Finger [Thu, 15 Dec 2016 18:23:01 +0000 (12:23 -0600)]
rtlwifi: rtl8723ae: Remove all instances of DBG_EMERG

This is a step toward eliminating the RT_TRACE macros. Those calls that
have DBG_EMERG as the level are always logged, and they represent error
conditions, thus they are replaced with pr_err().

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agortlwifi: rtl8723be: Remove all instances of DBG_EMERG
Larry Finger [Thu, 15 Dec 2016 18:23:00 +0000 (12:23 -0600)]
rtlwifi: rtl8723be: Remove all instances of DBG_EMERG

This is a step toward eliminating the RT_TRACE macros. Those calls that
have DBG_EMERG as the level are always logged, and they represent error
conditions, thus they are replaced with pr_err().

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agortlwifi: rtl8821ae: Remove all instances of DBG_EMERG
Larry Finger [Thu, 15 Dec 2016 18:22:59 +0000 (12:22 -0600)]
rtlwifi: rtl8821ae: Remove all instances of DBG_EMERG

This is a step toward eliminating the RT_TRACE macros. Those calls that
have DBG_EMERG as the level are always logged, and they represent error
conditions, thus they are replaced with pr_err().

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agortlwifi: Remove RT_TRACE messages that use DBG_EMERG
Larry Finger [Thu, 15 Dec 2016 18:22:58 +0000 (12:22 -0600)]
rtlwifi: Remove RT_TRACE messages that use DBG_EMERG

These messages are always logged and represent error conditions, thus
we can use pr_err().

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agortlwifi: Replace local debug macro RT_ASSERT
Larry Finger [Thu, 15 Dec 2016 18:22:57 +0000 (12:22 -0600)]
rtlwifi: Replace local debug macro RT_ASSERT

This macro can be replaced with WARN_ONCE. In addition to using a
standard debugging macro for these critical errors, we also get
a stack dump.

In rtl8821ae/hw.c, a senseless comment was removed, and an incorrect
indentation was fixed.

This patch also fixes two places in each of rtl8192ee, rtl8723be,
and rtl8821ae where the logical condition was incorrect.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agort2x00: add mutex to synchronize config and link tuner
Stanislaw Gruszka [Mon, 19 Dec 2016 10:52:57 +0000 (11:52 +0100)]
rt2x00: add mutex to synchronize config and link tuner

Do not perform mac80211 config and link_tuner at the same time,
this can possibly result in wrong RF or BBP configuration.

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agort2800: replace msleep() with usleep_range() on channel switch
Stanislaw Gruszka [Mon, 19 Dec 2016 10:52:56 +0000 (11:52 +0100)]
rt2800: replace msleep() with usleep_range() on channel switch

msleep(1) can sleep much more time then requested 1ms, this is not good
on channel switch, which we want to be performed fast (i.e. to make scan
faster). Replace msleep() with usleep_range(), which has much smaller
maximum sleeping time boundary.

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agort2800: replace mdelay by usleep on vco calibration.
Stanislaw Gruszka [Mon, 19 Dec 2016 10:52:55 +0000 (11:52 +0100)]
rt2800: replace mdelay by usleep on vco calibration.

This procedure can sleep, hence mdelay is not needed.

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agort2x00: merge agc and vco works with link tuner
Stanislaw Gruszka [Mon, 19 Dec 2016 10:52:54 +0000 (11:52 +0100)]
rt2x00: merge agc and vco works with link tuner

We need to perform different actions (AGC and VCO calibrations and VGC
tuning) periodically at different intervals. We don't need separate
works for those, we can use link tuner work and just check for proper
interval on it.

This fixes performing AGC and VCO calibration when scanning on STA
mode. We need to be on-channel to perform those calibrations.

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agort2800: perform VCO recalibration for RF5592 chip
Stanislaw Gruszka [Mon, 19 Dec 2016 10:52:53 +0000 (11:52 +0100)]
rt2800: perform VCO recalibration for RF5592 chip

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agort2800: warn if doing VCO recalibration for unknow RF chip
Stanislaw Gruszka [Mon, 19 Dec 2016 10:52:52 +0000 (11:52 +0100)]
rt2800: warn if doing VCO recalibration for unknow RF chip

Since we reset TX_PIN_CFG register, we have to finish recalibration.
Warn if this is not possible.

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agort2800: rename adjust_freq_offset function
Stanislaw Gruszka [Mon, 19 Dec 2016 10:52:51 +0000 (11:52 +0100)]
rt2800: rename adjust_freq_offset function

We have different modes of adjusting freq offset on different chips.
Call current adjustment similarly like vendor driver - mode1 .

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agort2800: set MAX_PSDU len according to remote STAs capabilities
Stanislaw Gruszka [Mon, 19 Dec 2016 10:52:50 +0000 (11:52 +0100)]
rt2800: set MAX_PSDU len according to remote STAs capabilities

MAX_LEN_CFG_MAX_PSDU specify maximum transmitted by HW AMPDU length
(0 - 8kB, 1 - 16kB, 2 - 32kB, 3 - 64kB). Set this option according to
remote stations capabilities (based on HT ampdu_factor). However limit
the value based our hardware TX capabilities as some chips can not send
more than 16kB (factor 1). Limit for all chips is currently 32kB
(factor 2), but perhaps for some chips this could be increased
to 64kB by setting drv_data->max_psdu to 3.

Since MAX_LEN_CFG_MAX_PSDU is global setting, on multi stations modes
(AP, IBSS, mesh) we limit according to less capable remote STA. We can
not set bigger value to speed up communication with some stations and
do not break communication with slow stations.

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agort2800: set minimum MPDU and PSDU lengths to sane values
Stanislaw Gruszka [Mon, 19 Dec 2016 10:52:49 +0000 (11:52 +0100)]
rt2800: set minimum MPDU and PSDU lengths to sane values

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agort2800: don't set ht parameters for non-aggregated frames
Stanislaw Gruszka [Mon, 19 Dec 2016 10:52:48 +0000 (11:52 +0100)]
rt2800: don't set ht parameters for non-aggregated frames

Do not set ampdu_density and ba_size for frames without AMPDU bit i.e.
frames that will not be aggregated to AMPDU.

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agort2800: make rx ampdu_factor depend on number of rx chains
Stanislaw Gruszka [Mon, 19 Dec 2016 10:52:47 +0000 (11:52 +0100)]
rt2800: make rx ampdu_factor depend on number of rx chains

Initalize max ampdu_factor supported by us based on rx chains, vendor
driver do the same.

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agoorinoco: Use shash instead of ahash for MIC calculations
Andrew Lutomirski [Mon, 12 Dec 2016 20:55:55 +0000 (12:55 -0800)]
orinoco: Use shash instead of ahash for MIC calculations

Eric Biggers pointed out that the orinoco driver pointed scatterlists
at the stack.

Fix it by switching from ahash to shash.  The result should be
simpler, faster, and more correct.

Cc: stable@vger.kernel.org # 4.9 only
Reported-by: Eric Biggers <ebiggers3@gmail.com>
Signed-off-by: Andy Lutomirski <luto@kernel.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agoadm80211: return an error if adm8211_alloc_rings() fails
Dan Carpenter [Wed, 7 Dec 2016 11:21:22 +0000 (14:21 +0300)]
adm80211: return an error if adm8211_alloc_rings() fails

We accidentally return success when adm8211_alloc_rings() fails but we
should preserve the error code.

Fixes: cc0b88cf5ecf ("[PATCH] Add adm8211 802.11b wireless driver")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agoiwlegacy: make il3945_mac_ops __ro_after_init
Johannes Berg [Wed, 7 Dec 2016 06:36:46 +0000 (07:36 +0100)]
iwlegacy: make il3945_mac_ops __ro_after_init

There's no need for this to be only __read_mostly, since
it's only written in a single way depending on the module
parameter, so that can be moved into the module's __init
function, and the ops can be __ro_after_init.

This is a little bit safer since it means the ops can't
be overwritten (accidentally or otherwise), which would
otherwise cause an arbitrary function or bad pointer to
be called.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Acked-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agomwifiex: Enable dynamic bandwidth signalling
Amitkumar Karwar [Fri, 16 Dec 2016 07:25:45 +0000 (12:55 +0530)]
mwifiex: Enable dynamic bandwidth signalling

Enable dynamic bandwidth signalling by setting the corresponding
bit in MAC control register.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agomwifiex: change width of MAC control variable
Amitkumar Karwar [Fri, 16 Dec 2016 07:25:44 +0000 (12:55 +0530)]
mwifiex: change width of MAC control variable

Firmware has started making use of reserved field.
Accordingly change curr_pkt_filter from u16 to u32.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agomwifiex: sdio: fix use after free issue for save_adapter
Amitkumar Karwar [Thu, 1 Dec 2016 13:53:31 +0000 (19:23 +0530)]
mwifiex: sdio: fix use after free issue for save_adapter

If we have sdio work requests received when sdio card reset is
happening, we may end up accessing older save_adapter pointer
later which is already freed during card reset.
This patch solves the problem by cancelling those pending requests.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
7 years agoath10k: enable advertising support for channel 169, 5Ghz
Mohammed Shafi Shajakhan [Tue, 27 Dec 2016 15:23:35 +0000 (20:53 +0530)]
ath10k: enable advertising support for channel 169, 5Ghz

Enable advertising support for channel 169, 5Ghz so that
based on the regulatory domain(country code) this channel
shall be active for use. For example in countries like India
this channel shall be available for use with latest regulatory updates

Signed-off-by: Mohammed Shafi Shajakhan <mohammed@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
7 years agoath10k: Remove passing unused argument for ath10k_mac_tx
Mohammed Shafi Shajakhan [Tue, 27 Dec 2016 13:32:35 +0000 (19:02 +0530)]
ath10k: Remove passing unused argument for ath10k_mac_tx

'ath10k_mac_tx' does not seems to use the per station table
entry pointer 'sta' (struct ieee80211_sta), hence remove passing
this unused argument

Signed-off-by: Mohammed Shafi Shajakhan <mohammed@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
7 years agoath10k: ignore configuring the incorrect board_id
Ryan Hsu [Thu, 22 Dec 2016 23:02:37 +0000 (15:02 -0800)]
ath10k: ignore configuring the incorrect board_id

With command to get board_id from otp, in the case of following

  boot get otp board id result 0x00000000 board_id 0 chip_id 0
  boot using board name 'bus=pci,bmi-chip-id=0,bmi-board-id=0"
  ...
  failed to fetch board data for bus=pci,bmi-chip-id=0,bmi-board-id=0 from
  ath10k/QCA6174/hw3.0/board-2.bin

The invalid board_id=0 will be used as index to search in the board-2.bin.

Ignore the case with board_id=0, as it means the otp is not carrying
the board id information.

Signed-off-by: Ryan Hsu <ryanhsu@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
7 years agoath10k: recal the txpower when removing interface
Ryan Hsu [Thu, 22 Dec 2016 22:31:46 +0000 (14:31 -0800)]
ath10k: recal the txpower when removing interface

The txpower is being recalculated when adding interface to make sure
txpower won't overshoot the spec, and when removing the interface,
the txpower should again to be recalculated to restore the correct value
from the active interface list.

Following is one of the scenario
vdev0 is created as STA and connected: txpower:23
vdev1 is created as P2P_DEVICE for control interface: txpower:0
vdev2 is created as p2p go/gc interface: txpower is 21

So the vdev2@txpower:21 will be set to firmware when vdev2 is created.
When we tear down the vdev2, the txpower needs to be recalculated to
re-set it to vdev0@txpower:23 as vdev0/vdev1 are the active interface.

ath10k_pci mac vdev 0 peer create 8c:fd:f0:01:62:98
ath10k_pci mac vdev_id 0 txpower 23
... (adding interface)
ath10k_pci mac vdev create 2 (add interface) type 1 subtype 3
ath10k_pci mac vdev_id 2 txpower 21
ath10k_pci mac txpower 21
... (removing interface)
ath10k_pci mac vdev 2 delete (remove interface)
ath10k_pci vdev 1 txpower 0
ath10k_pci vdev 0 txpower 23
ath10k_pci mac txpower 23

Signed-off-by: Ryan Hsu <ryanhsu@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
7 years agoath10k: support dev_coredump for crash dump
Arun Khandavalli [Wed, 21 Dec 2016 12:19:21 +0000 (14:19 +0200)]
ath10k: support dev_coredump for crash dump

Whenever firmware crashes, and both CONFIG_ATH10K_DEBUGFS and
CONFIG_ALLOW_DEV_COREDUMP are enabled, dump information about the crash via a
devcoredump device. Dump can be read from userspace for further analysis from:

/sys/class/devcoredump/devcd*/data

As until now we have provided the firmware crash dump file via fw_crash_dump
debugfs keep it still available but deprecate and a warning print that the user
should switch to using dev_coredump.

Future improvement would be not to depend on CONFIG_ATH10K_DEBUGFS, as there
might be systems which want to get the firmware crash dump but not enable
debugfs. How to handle memory consumption is also something which needs to be
taken into account.

Signed-off-by: Arun Khandavalli <akhandav@qti.qualcomm.com>
[kvalo@qca.qualcomm.com: rebase, fixes, improve commit log]
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
7 years agoath10k: fix incorrect txpower set by P2P_DEVICE interface
Ryan Hsu [Tue, 13 Dec 2016 22:55:19 +0000 (14:55 -0800)]
ath10k: fix incorrect txpower set by P2P_DEVICE interface

Ath10k reports the phy capability that supports P2P_DEVICE interface.

When we use the P2P supported wpa_supplicant to start connection, it'll
create two interfaces, one is wlan0 (vdev_id=0) and one is P2P_DEVICE
p2p-dev-wlan0 which is for p2p control channel (vdev_id=1).

ath10k_pci mac vdev create 0 (add interface) type 2 subtype 0
ath10k_add_interface: vdev_id: 0, txpower: 0, bss_power: 0
...
ath10k_pci mac vdev create 1 (add interface) type 2 subtype 1
ath10k_add_interface: vdev_id: 1, txpower: 0, bss_power: 0

And the txpower in per vif bss_conf will only be set to valid tx power when
the interface is assigned with channel_ctx.

But this P2P_DEVICE interface will never be used for any connection, so
that the uninitialized bss_conf.txpower=0 is assinged to the
arvif->txpower when interface created.

Since the txpower configuration is firmware per physical interface.
So the smallest txpower of all vifs will be the one limit the tx power
of the physical device, that causing the low txpower issue on other
active interfaces.

wlan0: Limiting TX power to 21 (24 - 3) dBm
ath10k_pci mac vdev_id 0 txpower 21
ath10k_mac_txpower_recalc: vdev_id: 1, txpower: 0
ath10k_mac_txpower_recalc: vdev_id: 0, txpower: 21
ath10k_pci mac txpower 0

This issue only happens when we use the wpa_supplicant that supports
P2P or if we use the iw tool to create the control P2P_DEVICE interface.

Signed-off-by: Ryan Hsu <ryanhsu@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
7 years agoath10k: fix potential memory leak in ath10k_wmi_tlv_op_pull_fw_stats()
Christian Lamparter [Thu, 29 Dec 2016 14:12:09 +0000 (16:12 +0200)]
ath10k: fix potential memory leak in ath10k_wmi_tlv_op_pull_fw_stats()

ath10k_wmi_tlv_op_pull_fw_stats() uses tb = ath10k_wmi_tlv_parse_alloc(...)
function, which allocates memory. If any of the three error-paths are
taken, this tb needs to be freed.

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
7 years agoliquidio: optimize reads from Octeon PCI console
Felix Manlunas [Fri, 30 Dec 2016 01:04:47 +0000 (17:04 -0800)]
liquidio: optimize reads from Octeon PCI console

Reads from Octeon PCI console are inefficient because before each read
operation, a dynamic mapping to Octeon DRAM is set up.  This patch replaces
the repeated setup of a dynamic mapping with a one-time setup of a static
mapping.

Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: Raghu Vatsavayi <raghu.vatsavayi@cavium.com>
Signed-off-by: Derek Chickles <derek.chickles@cavium.com>
Signed-off-by: Satanand Burla <satananda.burla@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: dsa: Implement ndo_get_phys_port_id
Florian Fainelli [Thu, 29 Dec 2016 22:20:56 +0000 (14:20 -0800)]
net: dsa: Implement ndo_get_phys_port_id

Implement ndo_get_phys_port_id() by returning the physical port number
of the switch this per-port DSA created network interface corresponds
to.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agonet: dev_weight: TX/RX orthogonality
Matthias Tafelmeier [Thu, 29 Dec 2016 20:37:21 +0000 (21:37 +0100)]
net: dev_weight: TX/RX orthogonality

Oftenly, introducing side effects on packet processing on the other half
of the stack by adjusting one of TX/RX via sysctl is not desirable.
There are cases of demand for asymmetric, orthogonal configurability.

This holds true especially for nodes where RPS for RFS usage on top is
configured and therefore use the 'old dev_weight'. This is quite a
common base configuration setup nowadays, even with NICs of superior processing
support (e.g. aRFS).

A good example use case are nodes acting as noSQL data bases with a
large number of tiny requests and rather fewer but large packets as responses.
It's affordable to have large budget and rx dev_weights for the
requests. But as a side effect having this large a number on TX
processed in one run can overwhelm drivers.

This patch therefore introduces an independent configurability via sysctl to
userland.

Signed-off-by: Matthias Tafelmeier <matthias.tafelmeier@gmx.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agostmmac: adding EEE to GMAC4
jpinto [Thu, 29 Dec 2016 17:10:27 +0000 (17:10 +0000)]
stmmac: adding EEE to GMAC4

This patch adds Energy Efficiency Ethernet to GMAC4.

Signed-off-by: Joao Pinto <jpinto@synopsys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agosctp: refactor sctp_datamsg_from_user
Marcelo Ricardo Leitner [Thu, 29 Dec 2016 17:53:28 +0000 (15:53 -0200)]
sctp: refactor sctp_datamsg_from_user

This patch refactors sctp_datamsg_from_user() in an attempt to make it
better to read and avoid code duplication for handling the last
fragment.

It also avoids doing division and remaining operations. Even though, it
should still operate similarly as before this patch.

Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMerge branch 'bnxt_en-updates'
David S. Miller [Thu, 29 Dec 2016 19:37:25 +0000 (14:37 -0500)]
Merge branch 'bnxt_en-updates'

Michael Chan says:

====================
bnxt_en: updates for net-next.

This patch series for net-next contains cleanups, new features and minor
fixes.  The driver specific busy polling code is removed to use busy
polling support in core networking.  Hardware RFS support is enhanced with
added ipv6 flows support and VF support.  A new scheme to allocate TX
rings from the firmware is implemented for newer chips and firmware.  Plus
some misc. cleanups, minor fixes, and to add the maintainer entry.  Please
review.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agoMAINTAINERS: Add bnxt_en maintainer info.
Michael Chan [Thu, 29 Dec 2016 17:13:44 +0000 (12:13 -0500)]
MAINTAINERS: Add bnxt_en maintainer info.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobnxt_en: Handle no aggregation ring gracefully.
Michael Chan [Thu, 29 Dec 2016 17:13:43 +0000 (12:13 -0500)]
bnxt_en: Handle no aggregation ring gracefully.

The current code assumes that we will always have at least 2 rx rings, 1
will be used as an aggregation ring for TPA and jumbo page placements.
However, it is possible, especially on a VF, that there is only 1 rx
ring available.  In this scenario, the current code will fail to initialize.
To handle it, we need to properly set up only 1 ring without aggregation.
Set a new flag BNXT_FLAG_NO_AGG_RINGS for this condition and add logic to
set up the chip to place RX data linearly into a single buffer per packet.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobnxt_en: Set default completion ring for async events.
Michael Chan [Thu, 29 Dec 2016 17:13:42 +0000 (12:13 -0500)]
bnxt_en: Set default completion ring for async events.

With the added support for the bnxt_re RDMA driver, both drivers can be
allocating completion rings in any order.  The firmware does not know
which completion ring should be receiving async events.  Add an
extra step to tell firmware the completion ring number for receiving
async events after bnxt_en allocates the completion rings.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobnxt_en: Implement new scheme to reserve tx rings.
Michael Chan [Thu, 29 Dec 2016 17:13:41 +0000 (12:13 -0500)]
bnxt_en: Implement new scheme to reserve tx rings.

In order to properly support TX rate limiting in SRIOV VF functions or
NPAR functions, firmware needs better control over tx ring allocations.
The new scheme requires the driver to reserve the number of tx rings
and to query to see if the requested number of tx rings is reserved.
The driver will use the new scheme when the firmware interface spec is
1.6.1 or newer.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobnxt_en: Add IPV6 hardware RFS support.
Michael Chan [Thu, 29 Dec 2016 17:13:40 +0000 (12:13 -0500)]
bnxt_en: Add IPV6 hardware RFS support.

Accept ipv6 flows in .ndo_rx_flow_steer() and support ETHTOOL_GRXCLSRULE
ipv6 flows.

Signed-off-by: Michael Chan <michael.chan@broadocm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobnxt_en: Assign additional vnics to VFs.
Michael Chan [Thu, 29 Dec 2016 17:13:39 +0000 (12:13 -0500)]
bnxt_en: Assign additional vnics to VFs.

Assign additional vnics to VFs whenever possible so that NTUPLE can be
supported on the VFs.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobnxt_en: Add new hardware RFS mode.
Michael Chan [Thu, 29 Dec 2016 17:13:38 +0000 (12:13 -0500)]
bnxt_en: Add new hardware RFS mode.

The existing hardware RFS mode uses one hardware RSS context block
per ring just to calculate the RSS hash.  This is very wasteful and
prevents VF functions from using it.  The new hardware mode shares
the same hardware RSS context for RSS placement and RFS steering.
This allows VFs to enable RFS.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobnxt_en: Refactor code that determines RFS capability.
Michael Chan [Thu, 29 Dec 2016 17:13:37 +0000 (12:13 -0500)]
bnxt_en: Refactor code that determines RFS capability.

Add function bnxt_rfs_supported() that determines if the chip supports
RFS.  Refactor the existing function bnxt_rfs_capable() that determines
if run-time conditions support RFS.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobnxt_en: Add function to get vnic capability.
Michael Chan [Thu, 29 Dec 2016 17:13:36 +0000 (12:13 -0500)]
bnxt_en: Add function to get vnic capability.

The new vnic RSS capability will enhance NTUPLE support, to be added
in subsequent patches.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobnxt_en: Refactor TPA code path.
Michael Chan [Thu, 29 Dec 2016 17:13:35 +0000 (12:13 -0500)]
bnxt_en: Refactor TPA code path.

Call tcp_gro_complete() in the common code path instead of the chip-
specific method.  The newer 5731x method is missing the call.

Signed-off-by: Michael Chan <michael.chan@broadcmo.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobnxt_en: Fix and clarify link_info->advertising.
Michael Chan [Thu, 29 Dec 2016 17:13:34 +0000 (12:13 -0500)]
bnxt_en: Fix and clarify link_info->advertising.

The advertising field is closely related to the auto_link_speeds field.
The former is the user setting while the latter is the firmware setting.
Both should be u16.  We should use the advertising field in
bnxt_get_link_ksettings because the auto_link_speeds field may not
be updated with the latest from the firmware yet.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
7 years agobnxt_en: Improve the IRQ disable sequence during shutdown.
Michael Chan [Thu, 29 Dec 2016 17:13:33 +0000 (12:13 -0500)]
bnxt_en: Improve the IRQ disable sequence during shutdown.

The IRQ is disabled by writing to the completion ring doorbell.  This
should be done before the hardware completion ring is freed for correctness.
The current code disables IRQs after all the completion rings are freed.

Fix it by calling bnxt_disable_int_sync() before freeing the completion
rings.  Rearrange the code to avoid forward declaration.

Signed-off-by: Michael Chan <michael.chan@broadocm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>