net/enic: fix name of classifiers hash table

The enic_clsf_init() function is called once per enic instance, but it
used a static name to create the hash table.  Consequently when using
more than one enic instance, there was a name collision which caused
errors:

EAL: memzone_reserve_aligned_thread_unsafe():
  memzone<RG_HT_enicpmd_clsf_hash> already exists
RING: Cannot reserve memory
HASH: memory allocation failed
PMD: rte_enic_pmd: Init of hash table for clsf failed.
  Flow director feature will not work

This patch changes the name to be unique per enic instance.

Fixes: fefed3d1e62c ("enic: new driver")

Signed-off-by: Nelson Escobar <neescoba@cisco.com>
Reviewed-by: John Daley <johndale@cisco.com>

net/mlx: fix build with icc

Compilation errors:
mlx4:
  drivers/net/mlx4/mlx4.c(5409): error #188:
  enumerated type mixed with another type
          priv->intr_handle.type = 0;
                                 ^
mlx5:
  drivers/net/mlx5/mlx5_rxq.c(282): error #188:
  enumerated type mixed with another type
        enum hash_rxq_type type = 0;
                                  ^
and more same type of error.
Fix these by assigning enum values rather than integer values to the enum
variables

Fixes: c4da6caa426d ("mlx4: handle link status interrupts")
Fixes: 198a3c339a8f ("mlx5: handle link status interrupts")
Fixes: 0d2186743d62 ("mlx5: manage all special flow types at once")
Fixes: 612ad38209f7 ("mlx5: fix hash Rx queue type in RSS mode")
Fixes: 083c2dd31776 ("mlx5: refactor special flows handling")

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>

net/ixgbe: remove useless assignment

The "end" variable is assigned and then two lines later is assigned a
different value, making the first assignment useless. Remove it.

Coverity issue: 13335
Fixes: cf4b4708a88a ("ixgbe: improve slow-path perf with vector scattered Rx")

Signed-off-by: Daniel Mrzyglod <danielx.t.mrzyglod@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>

net/e1000/base: fix build with gcc 6

Add the missing braces to the 'if' statements to fix the misleading
identation. This also fixes the following build errors when building
with gcc >= 6:

drivers/net/e1000/base/e1000_phy.c:4156:2:
error: this 'if' clause does not guard... [-Werror=misleading-indentation]
    if (locked)

drivers/net/e1000/base/e1000_phy.c:4158:3:
note: ...this statement, but the latter is misleadingly indented as if
it is guarded by the 'if'
    if (!ready)

drivers/net/e1000/base/e1000_phy.c:4221:2:
error: this 'if' clause does not guard... [-Werror=misleading-indentation]
    if (locked)

drivers/net/e1000/base/e1000_phy.c:4223:3:
note: ...this statement, but the latter is misleadingly indented as if
it is guarded by the 'if'
    if (!ready)

Signed-off-by: Markos Chandras <mchandras@suse.de>

net/ixgbe/base: update documentation for base code update

The ixgbe base driver was updated to version
cid-10g-shared-code.2016.04.12

The changes include:
Added sgmii link for X550.
Added mac link setup for X550a SFP and SFP+.
Added KR support for X550em_a.
Added new phy definitions for M88E1500.
Added support for the VLVF to be bypassed when adding/removing
a VFTA entry.
Added X550a flow control auto negotiation support.

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: define bit to enable crosstalk workaround

A work around for a new crosstalk erratum that causes link flap in
entry cages has been introduced. So this patch defines the bit in
NVM that will tell software if this work around is needed.

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: add flow control autoneg for X550a

This patch adds X550a flow control auto negotiation support.
ixgbe_setup_fc_x550a and ixgbe_fc_autoneg_X550a functions are
added to setup and enable flow control. MAC ops function pointer
fc_autoneg was added so that hardware specific fc autoneg functions
can be called from ixgbe_fc_enable_generic.

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: allow setting MAC anti spoofing per VF

Make ixgbe_set_mac_anti_spoofing() consistent with the other
functions that deal with setting VLAN and Ethertype spoofing by
changing the prototype to accept a VF parameter.

Also change the logic for writing the PFVFSPOOF register to be similar
to the MAC and Ethertype functions.

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: fix endianness

This patch fixes endianness issues about host interface command.

Fixes: ad66a85dce9a ("ixgbe/base: new FW values")
Fixes: 0790adeb5675 ("ixgbe/base: support X550em_a device")

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: use 8 bits for PCI functions

Since PCIe standard defines maximum of 8 functions per device lan_id
is a value 0..7. Because of that, lan_id don't need to be u16.

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: unify coding style

This patch changes static keyword to STATIC definition, which can be
redefined depending on the compiler used.

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: allow bypassing VLAN pool filters

This patch adds support for the VLAN pool filter (VLVF) to be
bypassed when adding or removing a VLAN filter table array (VFTA) entry.
The PF can utilize the default pool while preserving the VLVF for the
VFs use.
Meanwhile, update the VF operations and drivers where corresponding
functionality is invoked.

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: simplify VLAN management

This patch simplifies the adding and removing VLANs from
VFTA/VLVF/VLVFB registers. The logic to determine the registers to use
has been simplified to (vid / 32) and (1 - vid / 32). Many conditional
paths and checks are no longer needed with this patch.

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: limit PHY token lock to MDIO access

This patch limits getting and putting the PHY Token to PHY MDIO
access only by adding ixgbe_read_phy_reg_x550a and
ixgbe_write_phy_reg_x550a. The PHY Token is only needed to
synchronize access to the MDIO shared between the two MAC instance.

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: fix register access error

This patch corrects the FLA/GSCL/GSCN access offset values according
to the datasheet.

Fixes: 0790adeb5675 ("ixgbe/base: support X550em_a device")

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: fix race condition between ports

This patch fixes a possible race issue between ports, when issuing host
interface commands, by acquiring/releasing the management host interface
semaphore in ixgbe_host_interface_command.

Fixes: 36f43e8679ae ("ixgbe/base: refactor manageability block communication")

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: add function to reset swfw semaphore

For X540 onwards it is possible if a system reset occurs at the
right time to leave the SWFW semaphore high. This new function will
attempt to grab and release the semaphore. If the grab times out it
will still release the semaphore placing it in a known good state.
The idea is to call this when you know no one should be holding the
semaphore (i.e. probe time)

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: update device IDs

There are two device IDs changed from 15C6/15C7 to 15E4/15E5 because of
PHY info changes. 15C6/15C7 IDs are now used for the backplane
SGMII versions.
Also, clean up some discovery kludges from the previous shared ID,
and also add 15C6/15C7 to ixgbe_set_mdio_speed just for paranoia
to control MDIO speed even though nothing should be attached.

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: support new PHY types

This patch adds new phy definitions and support functions to enable those
new phy types.

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: refactor internal PHY mode determination

NW_MNG_IF_SEL register is a PHY link configuration register.
Add ixgbe_read_mng_if_sel_x550em to read NW_MNG_IF_SEL, validate
register values and save fields such as PHY MDIO address. This
centralises the reading and checking of the register in one place

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: rename header guard

The ixgbe_vf.h file did not use _<FILENAME>_ and instead used
__<FILENAME>__ which is not the standard used in every other file.

Fixes: af75078fece3 ("first public release")

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: fix error path to release lock

When there is an error getting the PHY token, the error path
fails to release the locks that it has taken. Release those
locks in that failure case.

Fixes: 86b8fb293fdf ("ixgbe/base: add sw-firmware sync for resource sharing on X550em_a")

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: fix macro name

This patch renames IXGBE_PVFTTDLEN to IXGBE_PVFTDLEN according to
abbreviation of Transmit Descriptor Length in datasheet.

Fixes: d2e72774e58c ("ixgbe/base: support X550")

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: clear stale VMDq pool mappings

This patch adds clearing the pool mappings when configuring default
MAC addresses for the interface. Without this there will be the risk
of leaking an address into pool 0 which really belongs to VF 0 when
SR-IOV is enabled.

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: add link MAC setup for X550a SFP+

This patch updates ixgbe_setup_mac_link_sfp_x550a for X550 SFP+.
ixgbe_set_lan_id_multi_port_pcie has been updated to set the MAC
instance(0/1) which is needed when configuring the external PHY,
since X550a has two instances of MGPK. The MAC instance is read
from the EEPROM.

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: allow per-device-type method calls

Use the method pointers instead of direct function calls for IOSF
access so that the right functions can be called on X550EM_a,
compared to other devices using the driver.

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: add KR support for X550em_a devices

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: refactor EEE setup for X550

Break ixgbe_setup_eee_X550 down to better handle a change from if
statements to switch statements needed to add X550em_a KR support.

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: fix firmware command checksum error

When software sends commands to firmware using the host
slave command interface, firmware fails to receive the
command due to a checksum failed error, as the checksum is
not being correctly set by the driver software.

This patch sets command checksum to the default value of
0xFF, as per the datasheet, therefore the checksum won't
be checked by firmware.

Fixes: 86b8fb293fdf ("ixgbe/base: add sw-firmware sync for resource sharing on X550em_a")
Fixes: 0790adeb5675 ("ixgbe/base: support X550em_a device")

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: add MAC link setup for X550a SFP

This patch adds ixgbe_setup_mac_link_sfp_x550a for X550a SFP.

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: return error on VF MAC address failure

An error code indicating that the PF rejects the MAC address change
should be returned, in case that the PF has already assigned a MAC
for the VF.

Fixes: af75078fece3 ("first public release")

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: add SGMII link for X550

This patch adds new phy type and media type to support
SGMII link for X550, and add ixgbe_setup_sgmii to support
SGMII link setup.

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/ixgbe/base: add new VF requests to mailbox API

This patch adds two new VF requests of IXGBE_VF_GET_RETA and
IXGBE_VF_GET_RSS_KEY to the mailbox API.

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/mlx5: re-add Rx scatter support

This commit brings back Rx scatter and related support by the MTU update
function. The maximum number of segments per packet is not a fixed value
anymore (previously MLX5_PMD_SGE_WR_N, set to 4 by default) as it caused
performance issues when fewer segments were actually needed as well as
limitations on the maximum packet size that could be received with the
default mbuf size (supporting at most 8576 bytes).

These limitations are now lifted as the number of SGEs is derived from the
MTU (which implies MRU) at queue initialization and during MTU update.

Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Vasily Philipov <vasilyf@mellanox.com>
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>

net/mlx5: make Rx queue reinitialization safer

The primary purpose of rxq_rehash() function is to stop and restart
reception on a queue after re-posting buffers. This may fail if the array
that temporarily stores existing buffers for reuse cannot be allocated.

Update rxq_rehash() to work on the target queue directly (not through a
template copy) and avoid this allocation.

rxq_alloc_elts() is modified accordingly to take buffers from an existing
queue directly and update their refcount.

Unlike rxq_rehash(), rxq_setup() must work on a temporary structure but
should not allocate new mbufs from the pool while reinitializing an
existing queue. This is achieved by using the refcount-aware
rxq_alloc_elts() before overwriting queue data.

Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Vasily Philipov <vasilyf@mellanox.com>

net/mlx5: remove redundant Rx queue initialization

Toggling RX checksum offloads is already done at initialization time. This
code does not belong in rxq_rehash().

Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>

net/mlx5: re-add Tx gather support

Compared to its previous incarnation, the software limit on the number of
mbuf segments is no more (previously MLX5_PMD_SGE_WR_N, set to 4 by
default) hence no need for linearization code and related buffers that
permanently consumed a non negligible amount of memory to handle oversized
mbufs.

The resulting code is both lighter and faster.

With the addition of this code, older GCC versions (such
as 4.8.5) may complain about 'wqe' variable being uninitialized, so
initialize it preemptively, even though it is not necessary to do so.

Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>

net/mlx5: check remaining space while processing Tx burst

The space necessary to store segmented packets cannot be known in advance
and must be verified for each of them.

Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>

net/mlx5: add debug information about Tx queues capabilities

Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>

net/mlx5: support multi-packet send

This feature enables the TX burst function to emit up to 5 packets using
only two work queue entries (WQEs) on devices that support it. Saves PCI
bandwidth and improves performance.

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Olga Shern <olgas@mellanox.com>

net/mlx5: support inline send

Implement send inline feature which copies packet data directly into
work queue entries (WQEs) for improved latency. The maximum packet
size and the minimum number of Tx queues to qualify for inline send
are user-configurable.

This feature is effective when HW causes a performance bottleneck.

Signed-off-by: Yaacov Hazan <yaacovh@mellanox.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>

net/mlx5: replace countdown with threshold for Tx completions

Replacing the variable countdown (which depends on the number of
descriptors) with a fixed relative threshold known at compile time improves
performance by reducing the TX queue structure footprint and the amount of
code to manage completions during a burst.

Completions are now requested at most once per burst after threshold is
reached.

Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Vasily Philipov <vasilyf@mellanox.com>

net/mlx5: handle Rx CQE compression

Mini (compressed) completion queue entries (CQEs) are returned by the
NIC when PCI back pressure is detected, in which case the first CQE64
contains common packet information followed by a number of CQE8
providing the rest, followed by a matching number of empty CQE64
entries to be used by software for decompression.

Before decompression:

      0           1          2           6         7         8
  +-------+  +---------+ +-------+   +-------+ +-------+ +-------+
  | CQE64 |  |  CQE64  | | CQE64 |   | CQE64 | | CQE64 | | CQE64 |
  |-------|  |---------| |-------|   |-------| |-------| |-------|
  | ..... |  | cqe8[0] | |       | . |       | |       | | ..... |
  | ..... |  | cqe8[1] | |       | . |       | |       | | ..... |
  | ..... |  | ....... | |       | . |       | |       | | ..... |
  | ..... |  | cqe8[7] | |       |   |       | |       | | ..... |
  +-------+  +---------+ +-------+   +-------+ +-------+ +-------+

After decompression:

      0          1     ...     8
  +-------+  +-------+     +-------+
  | CQE64 |  | CQE64 |     | CQE64 |
  |-------|  |-------|     |-------|
  | ..... |  | ..... |  .  | ..... |
  | ..... |  | ..... |  .  | ..... |
  | ..... |  | ..... |  .  | ..... |
  | ..... |  | ..... |     | ..... |
  +-------+  +-------+     +-------+

This patch does not perform the entire decompression step as it would be
really expensive, instead the first CQE64 is consumed and an internal
context is maintained to interpret the following CQE8 entries directly.

Intermediate empty CQE64 entries are handed back to HW without further
processing.

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Olga Shern <olgas@mellanox.com>
Signed-off-by: Vasily Philipov <vasilyf@mellanox.com>

net/mlx5: refactor Tx data path

Bypass Verbs to improve Tx performance.

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Yaacov Hazan <yaacovh@mellanox.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>

net/mlx5: refactor Rx data path

Bypass Verbs to improve RX performance.

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Yaacov Hazan <yaacovh@mellanox.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Vasily Philipov <vasilyf@mellanox.com>

net/mlx5: add Tx/Rx burst function selection wrapper

These wrappers are meant to prevent code duplication later.

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>

net/mlx5: add support for configuration through kvargs

The intent is to replace the remaining compile-time options and environment
variables with a common mean of runtime configuration. This commit only
adds the kvargs handling code, subsequent commits will update the rest.

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>

net/mlx5: add definitions for data path without Verbs

These structures and macros extend those exposed by libmlx5 (in mlx5_hw.h)
to let the PMD manage work queue and completion queue elements directly.

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>

net/mlx5: update prerequisites for upcoming enhancements

The latest version of Mellanox OFED exposes hardware definitions necessary
to implement data path operation bypassing Verbs. Update the minimum
version requirement to MLNX_OFED >= 3.3 and clean up compatibility checks
for previous releases.

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>

net/mlx5: split Rx queue structure

To keep the data path as efficient as possible, move fields only useful to
the control path into new structure rxq_ctrl.

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>

net/mlx5: split Tx queue structure

To keep the data path as efficient as possible, move fields only useful to
the control path into new structure txq_ctrl.

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>

net/mlx5: remove inline Tx support

Inline TX will be fully managed by the PMD after Verbs is bypassed in the
data path. Remove the current code until then.

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>

net/mlx5: remove configuration variable

There is no scatter/gather support anymore, CONFIG_RTE_LIBRTE_MLX5_SGE_WR_N
has no purpose and can be removed.

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>

net/mlx5: remove Rx scatter support

This is done in preparation of bypassing Verbs entirely for the data path
as a performance improvement. RX scatter cannot be maintained during the
transition and will be reimplemented later.

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>

net/mlx5: remove Tx gather support

This is done in preparation of bypassing Verbs entirely for the data path
as a performance improvement. TX gather cannot be maintained during the
transition and will be reimplemented later.

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>

net/mlx5: split memory registration function

Except for the first time when memory registration occurs, the lkey is
always cached. Since memory registration is slow and performs system calls,
performance can be improved by moving that code to its own function outside
of the data path so only the lookup code is left in the original inlined
function.

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>

net: fix PCI class id

Use RTE_PCI_DEVICE macro to set all fields rather than explicitly setting
them individually in the code. This shortens the code while helping to
future-proof against future changes to the rte_pci_id structure.

Fixes: 701c8d80c820 ("pci: support class id probing")

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>

net/vmxnet3: fix dump format

Remove 0x prefix for %p format to prevent double 0x in logs

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: Yong Wang <yongwang@vmware.com>

doc: update UIO module in nfp guide

Removing dependency on nfp_uio kernel module. The igb_uio
kernel modules can be used instead.

Fixes: 80bc1752f16e ("nfp: add guide")

Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
Acked-by: John McNamara <john.mcnamara@intel.com>

net/ixgbe: fix single VLAN tag to be outer VLAN tag

Previously, a single VLAN header is treated as inner VLAN,
but generally, a single VLAN header is treated as the outer
VLAN header.
The patch fixes the ether type of a single VLAN type, and
enables configuring inner and outer TPID for double VLAN.

Fixes: 19b16e2f6442 ("ethdev: add vlan type when setting ether type")

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/i40e: fix single VLAN tag to be outer VLAN tag

In current i40e codebase, if single VLAN header is added in a packet,
it's treated as inner VLAN. Generally, a single VLAN header is
treated as the outer VLAN header, so update the driver behaviour
appropriately.

Fixes: 19b16e2f6442 ("ethdev: add vlan type when setting ether type")

Signed-off-by: Beilei Xing <beilei.xing@intel.com>

net/bnxt: support Cumulus+ Ethernet adapter

This patch adds support for Cumulus+ Ethernet adapters.
These Cumulus+ Ethernet adapters support 10Gb/25Gb/40Gb/50Gb speeds.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>

net/enic: fix negative array index write

Negative array index write using variable pos as an index to array
enic->fdir.nodes. Fixed by add array index check.

Coverity issue: 13270

Fixes: fefed3d1e62c ("enic: new driver")

Signed-off-by: John Daley <johndale@cisco.com>

net/ixgbe: use mbuf prefetch function

made second cache line access behavior same as IA

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Reviewed-by: Jianbo Liu <jianbo.liu@linaro.org>

net/bnx2x: fix incorrect buffer length on Rx

The Rx function should not be setting the mbuf buffer length, so remove
the assignment.

Fixes: 540a211084a7 ("bnx2x: driver core")

Signed-off-by: Charles (Chas) Williams <ciwillia@brocade.com>
Acked-by: Harish Patil <harish.patil@qlogic.com>

net/enic: support scattered Rx

For performance reasons, this patch uses 2 VIC RQs per RQ presented to
DPDK.

The VIC requires that each descriptor be marked as either a start of
packet (SOP) descriptor or a non-SOP descriptor.  A one RQ solution
requires skipping descriptors when receiving small packets and results
in bad performance when receiving many small packets.

The 2 RQ solution makes use of the VIC feature that allows a receive
on primary queue to 'spill over' into another queue if the receive is
too large to fit in the buffer assigned to the descriptor on the
primary queue.  This means that there is no skipping of descriptors
when receiving small packets and results in much better performance.

Signed-off-by: Nelson Escobar <neescoba@cisco.com>
Reviewed-by: John Daley <johndale@cisco.com>

net/e1000: configure outer VLAN TPID field

This patch enables configuring the outer TPID for double VLAN.
Note that all other TPID values, for single VLANs or inner VLAN in the
QinQ case, are read only.

Signed-off-by: Beilei Xing <beilei.xing@intel.com>
Acked-by: Helin Zhang <helin.zhang@intel.com>

net/qede: update version to 8.7.9.0_1.1.0.1

Signed-off-by: Harish Patil <harish.patil@qlogic.com>

net/qede: enable VF-VF traffic with unmatched dest address

This patch enables VF to VF traffic with unmatched destination addresses.
The steps to enable this are:
- Enable promiscuous mode filter settings.
- Check for VF mode and enable promiscuous mode settings for VF.
- Check filter configuration to ensure conflicting filter modes
are not set.

Signed-off-by: Sony Chacko <sony.chacko@qlogic.com>

net/qede: support 100G

 - Add device id to the PCI table
 - Add polling for the slowpath events for CMT mode device
 - Add prerequisites to allow 100g mode
        * Min number of queues needed is 2
        * Only even number of queues are allowed
 - Update documentation

Signed-off-by: Harish Patil <harish.patil@qlogic.com>

net/qede: set MTU

Signed-off-by: Sony Chacko <sony.chacko@qlogic.com>
[NIC overview doc update]
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>

net/qede: query RSS redirection table

Signed-off-by: Sony Chacko <sony.chacko@qlogic.com>

net/qede: update RSS redirection table

Add support for configuring the RSS redirection table and update
corresponding documentation.

Signed-off-by: Sony Chacko <sony.chacko@qlogic.com>

net/qede: get RSS hash configuration

Signed-off-by: Sony Chacko <sony.chacko@qlogic.com>

net/qede: support RSS hash configuration

Add support for setting hash configuration based on adapter capability
and update corresponding NIC documentation.

Signed-off-by: Sony Chacko <sony.chacko@qlogic.com>

net/ixgbe: fix missed packet types on Rx

ixgbe PMD RX function(s) misses some packet types that are:
 - correctly recognised by the underlying HW.
 - marked as supported by ixgbe_dev_supported_ptypes_get().

Fixes: 9586ebd358d5 ("ixgbe: replace some offload flags with packet type")

Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Tested-by: Olivier Matz <olivier.matz@6wind.com>

doc: remove Match Interface in fm10k guide

Intel stopped supporting Match Interface, remove reference to it in the
documentation.

Signed-off-by: Chen Jing D(Mark) <jing.d.chen@intel.com>
Acked-by: John McNamara <john.mcnamara@intel.com>

net/enic: fix releasing mbufs when tearing down Rx queue

When trying to release the mbufs, the function was incorrectly
iterating over the max size configured instead of the actual size
of the ring.

Fixes: 947d860c821f ("enic: improve Rx performance")

Signed-off-by: Nelson Escobar <neescoba@cisco.com>
Reviewed-by: John Daley <johndale@cisco.com>

net/enic: improve packet type identification

- add l4 ptypes to the ones we report as supporting
- report/use RTE_PTYPE_L3_IPV4_EXT_UNKNOWN and
  RTE_PTYPE_L3_IPV6_EXT_UNKNOWN instead of RTE_PTYPE_L3_IPV4 and
  RTE_PTYPE_L3_IPV6 as vic can't distinguish between packets with
  extentions and those without extentions.
- correctly set the ptype bits set on packets that are both tcp/udp
  and a frag
- set RTE_PTYPE_L4_NONFRAG on ip packets we know are not udp, tcp,
  or fragments.

Fixes: 947d860c821f ("enic: improve Rx performance")

Signed-off-by: Nelson Escobar <neescoba@cisco.com>
Reviewed-by: John Daley <johndale@cisco.com>

net/i40e: fix offload flags for vector Rx

The flags for RSS and flow director are not set correctly in the vector
Rx function, so applications which use these flags will not work
correctly.

The problem is caused by incorrect constants for masking,
shuffling and shifting the descriptor bytes, to create the resultant
flags in the mbuf. Correcting the constants fixes the problem

Fixes: 9ed94e5bb04e ("i40e: add vector Rx")

Signed-off-by: Zhe Tao <zhe.tao@intel.com>
Reviewed-by: Piotr Azarewicz <piotrx.t.azarewicz@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>

net/xenvirt: support dynamic page size

PAGE_SIZE constant is not defined on ARM since multiple values
are possible, so DPDK needs to dynamically get the page size.

Signed-off-by: Ricardo Salveti <ricardo.salveti@linaro.org>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>

net/bnx2x: fix MSIX vector and VF resource counts

If MSIX is available, the vector count given by the table size is one
less than the actual count.  This count also limits the receive and
transmit queue resources the VF can support.

Fixes: 540a211084a7 ("bnx2x: driver core")

Signed-off-by: Charles (Chas) Williams <ciwillia@brocade.com>
Acked-by: Harish Patil <harish.patil@qlogic.com>

app/test: add external state machine in mode 4

This adds test cases for exercising the external state machine API to
the mode 4 autotest.

Signed-off-by: Eric Kinzie <ehkinzie@gmail.com>
Acked-by: Declan Doherty <declan.doherty@intel.com>

net/bonding: allow external state machine in mode 4

Provide functions to allow an external 802.3ad state machine to transmit
and receive LACPDUs and to set the collection/distribution flags on
slave interfaces.

Signed-off-by: Eric Kinzie <ehkinzie@gmail.com>
Acked-by: Declan Doherty <declan.doherty@intel.com>

net/bonding: inherit maximum Rx packet length

Instead of a hard-coded maximum receive length, allow the bonded interface
to inherit this limit from the slave interfaces.  This allows
an application that uses jumbo frames to pass realistic values to
rte_eth_dev_configure without causing an error.

Before the bonding interface is configured, allow slaves with any
max_rx_pktlen to be added and remember the lowest of these values as
a candidate value.  During dev_configure, set the bond device's
max_rx_pktlen to the candidate value.  After this point only slaves
with a max_rx_pktlen greater or equal to that of the bonding device
can be added.

If all slaves are removed, the bond device's pktlen is cleared.

Signed-off-by: Eric Kinzie <ehkinzie@gmail.com>
Acked-by: Declan Doherty <declan.doherty@intel.com>

doc: add ThunderX nicvf

Updated doc/guides/nics/overview.rst, doc/guides/nics/thunderx.rst
and release notes

Changed "*" to "P" in overview.rst to capture the partially supported
feature as "*" creating alignment issues with Sphinx table

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Signed-off-by: Slawomir Rosek <slawomir.rosek@semihalf.com>
Signed-off-by: Maciej Czekaj <maciej.czekaj@caviumnetworks.com>
Acked-by: John McNamara <john.mcnamara@intel.com>

net/thunderx: add device start/stop and close

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Signed-off-by: Maciej Czekaj <maciej.czekaj@caviumnetworks.com>
Signed-off-by: Kamil Rytarowski <kamil.rytarowski@caviumnetworks.com>
Signed-off-by: Zyta Szpak <zyta.szpak@semihalf.com>
Signed-off-by: Slawomir Rosek <slawomir.rosek@semihalf.com>
Signed-off-by: Radoslaw Biernacki <rad@semihalf.com>

net/thunderx: support Tx queue start and stop

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Signed-off-by: Maciej Czekaj <maciej.czekaj@caviumnetworks.com>
Signed-off-by: Kamil Rytarowski <kamil.rytarowski@caviumnetworks.com>
Signed-off-by: Zyta Szpak <zyta.szpak@semihalf.com>
Signed-off-by: Slawomir Rosek <slawomir.rosek@semihalf.com>
Signed-off-by: Radoslaw Biernacki <rad@semihalf.com>

net/thunderx: support Rx queue start and stop

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Signed-off-by: Maciej Czekaj <maciej.czekaj@caviumnetworks.com>
Signed-off-by: Kamil Rytarowski <kamil.rytarowski@caviumnetworks.com>
Signed-off-by: Zyta Szpak <zyta.szpak@semihalf.com>
Signed-off-by: Slawomir Rosek <slawomir.rosek@semihalf.com>
Signed-off-by: Radoslaw Biernacki <rad@semihalf.com>

net/thunderx: get Rx queue count

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Signed-off-by: Maciej Czekaj <maciej.czekaj@caviumnetworks.com>
Signed-off-by: Kamil Rytarowski <kamil.rytarowski@caviumnetworks.com>
Signed-off-by: Zyta Szpak <zyta.szpak@semihalf.com>
Signed-off-by: Slawomir Rosek <slawomir.rosek@semihalf.com>
Signed-off-by: Radoslaw Biernacki <rad@semihalf.com>

net/thunderx: query supported packet types

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Signed-off-by: Maciej Czekaj <maciej.czekaj@caviumnetworks.com>
Signed-off-by: Kamil Rytarowski <kamil.rytarowski@caviumnetworks.com>
Signed-off-by: Zyta Szpak <zyta.szpak@semihalf.com>
Signed-off-by: Slawomir Rosek <slawomir.rosek@semihalf.com>
Signed-off-by: Radoslaw Biernacki <rad@semihalf.com>

net/thunderx: add single and multi-segment Rx

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Signed-off-by: Maciej Czekaj <maciej.czekaj@caviumnetworks.com>
Signed-off-by: Kamil Rytarowski <kamil.rytarowski@caviumnetworks.com>
Signed-off-by: Zyta Szpak <zyta.szpak@semihalf.com>
Signed-off-by: Slawomir Rosek <slawomir.rosek@semihalf.com>
Signed-off-by: Radoslaw Biernacki <rad@semihalf.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/thunderx: add single and multi-segment Tx

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Signed-off-by: Maciej Czekaj <maciej.czekaj@caviumnetworks.com>
Signed-off-by: Kamil Rytarowski <kamil.rytarowski@caviumnetworks.com>
Signed-off-by: Zyta Szpak <zyta.szpak@semihalf.com>
Signed-off-by: Slawomir Rosek <slawomir.rosek@semihalf.com>
Signed-off-by: Radoslaw Biernacki <rad@semihalf.com>

net/thunderx: add statistics

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Signed-off-by: Maciej Czekaj <maciej.czekaj@caviumnetworks.com>
Signed-off-by: Kamil Rytarowski <kamil.rytarowski@caviumnetworks.com>
Signed-off-by: Zyta Szpak <zyta.szpak@semihalf.com>
Signed-off-by: Slawomir Rosek <slawomir.rosek@semihalf.com>
Signed-off-by: Radoslaw Biernacki <rad@semihalf.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/thunderx: support promiscuous mode

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Signed-off-by: Maciej Czekaj <maciej.czekaj@caviumnetworks.com>
Signed-off-by: Kamil Rytarowski <kamil.rytarowski@caviumnetworks.com>
Signed-off-by: Zyta Szpak <zyta.szpak@semihalf.com>
Signed-off-by: Slawomir Rosek <slawomir.rosek@semihalf.com>
Signed-off-by: Radoslaw Biernacki <rad@semihalf.com>

net/thunderx: support MTU configuration

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Signed-off-by: Maciej Czekaj <maciej.czekaj@caviumnetworks.com>
Signed-off-by: Kamil Rytarowski <kamil.rytarowski@caviumnetworks.com>
Signed-off-by: Zyta Szpak <zyta.szpak@semihalf.com>
Signed-off-by: Slawomir Rosek <slawomir.rosek@semihalf.com>
Signed-off-by: Radoslaw Biernacki <rad@semihalf.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/thunderx: support RSS and RETA query and update

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Signed-off-by: Maciej Czekaj <maciej.czekaj@caviumnetworks.com>
Signed-off-by: Kamil Rytarowski <kamil.rytarowski@caviumnetworks.com>
Signed-off-by: Zyta Szpak <zyta.szpak@semihalf.com>
Signed-off-by: Slawomir Rosek <slawomir.rosek@semihalf.com>
Signed-off-by: Radoslaw Biernacki <rad@semihalf.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/thunderx: add Tx queue setup and release

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Signed-off-by: Maciej Czekaj <maciej.czekaj@caviumnetworks.com>
Signed-off-by: Kamil Rytarowski <kamil.rytarowski@caviumnetworks.com>
Signed-off-by: Zyta Szpak <zyta.szpak@semihalf.com>
Signed-off-by: Slawomir Rosek <slawomir.rosek@semihalf.com>
Signed-off-by: Radoslaw Biernacki <rad@semihalf.com>

net/thunderx: add Rx queue setup and release

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Signed-off-by: Maciej Czekaj <maciej.czekaj@caviumnetworks.com>
Signed-off-by: Kamil Rytarowski <kamil.rytarowski@caviumnetworks.com>
Signed-off-by: Zyta Szpak <zyta.szpak@semihalf.com>
Signed-off-by: Slawomir Rosek <slawomir.rosek@semihalf.com>
Signed-off-by: Radoslaw Biernacki <rad@semihalf.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/thunderx: get device infos

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Signed-off-by: Maciej Czekaj <maciej.czekaj@caviumnetworks.com>
Signed-off-by: Kamil Rytarowski <kamil.rytarowski@caviumnetworks.com>
Signed-off-by: Zyta Szpak <zyta.szpak@semihalf.com>
Signed-off-by: Slawomir Rosek <slawomir.rosek@semihalf.com>
Signed-off-by: Radoslaw Biernacki <rad@semihalf.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>