event/octeontx2: support xstats

Add support for retrieving statistics from SSO GWS and GGRP.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>

event/octeontx2: add register dump functions

Add SSO GWS and GGRP register dump function to aid debugging.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>

event/octeontx2: add SSO GWS and GGRP IRQ handlers

Register and implement SSO GWS and GGRP IRQ handlers for error
interrupts.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Signed-off-by: Jerin Jacob <jerinj@marvell.com>

event/octeontx2: support dequeue timeout tick conversion

Add function to convert dequeue timeout from ns to ticks.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>

event/octeontx2: support linking queues to ports

Links between queues and ports are controlled by setting/clearing GGRP
membership in SSOW_LF_GWS_GRPMSK_CHG.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>

event/octeontx2: add port config functions

Add default config, setup and release functions for event ports
i.e. SSO GWS.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>

event/octeontx2: add devargs for inflight buffer count

The number of events for a *open system* event device is specified
as -1 as per the eventdev specification.
Since, Octeontx2 SSO inflight events are only limited by DRAM size, the
xae_cnt devargs parameter is introduced to provide upper limit for
in-flight events.

Example:
--dev "0002:0e:00.0,xae_cnt=8192"

Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>

event/octeontx2: allocate event inflight buffers

Allocate buffers in DRAM that hold inflight events.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>

event/octeontx2: add event queue config functions

Add default config, setup and release functions for event queues i.e.
SSO GGRPS.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>

event/octeontx2: add device configure function

Add the device configure function that attaches the requested number of
SSO GWS(event ports) and GGRP(event queues) LF's to the PF.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>

event/octeontx2: add device capabilities function

Add the info_get function to return details on the queues, flow,
prioritization capabilities, etc. which this device has.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>

event/octeontx2: add init and fini for SSO object

SSO object needs to be initialized to communicate with the kernel AF
driver through mbox using the common API's.
Also, initialize the internal eventdev structure to defaults.
Attach NPA lf to the PF if needed.

Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>

event/octeontx2: add build infra and device probe

Add the make and meson based build infrastructure along with the
eventdev(SSO) device probe.

Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>

test/eventdev: fix producer core validity checks

When producer type is event timer adapter producer lcore checks are
skipped. Since, timer adapter relies on SW to arm timers producer lcore
is essential for its functionality.
Verify producer lcore validity when producer type is event timer
adapter.

Fixes: b01974da9f25 ("app/eventdev: add ethernet device producer option")
Cc: stable@dpdk.org
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>

app/eventdev: fix order test port creation

Configure event ports based on the underlying event device info rather
than using hardcoded values.

Fixes: 5710e751813e ("app/testeventdev: add order port setup")
Cc: stable@dpdk.org
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>

eventdev: fix doxygen comment

Update rte_eventdev.h file description to adapt to Tx adapter changes.

Fixes: c9bf83947e2e ("eventdev: add eth Tx adapter APIs")
Cc: stable@dpdk.org
Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Reviewed-by: Nikhil Rao <nikhil.rao@intel.com>

eventdev: change Rx adapter callback and stats structure

Replace the mbuf pointer array in the event eth Rx adapter
callback with an event array. Using an event array allows
the application to change attributes of the events enqueued
by the SW adapter.

The callback can drop packets and populate a callback
argument with the number of dropped packets. Add a Rx adapter
stats field to keep track of the total number of dropped packets.

This commit removes the experimental tags from
the callback and stats APIs, the experimental tag from eventdev
is also removed and eventdev functions become part of the
main DPDK API/ABI.

Signed-off-by: Nikhil Rao <nikhil.rao@intel.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>

event/dpaa2: fix timeout ticks

Correct timeout to tick conversion.

Fixes: 0ce3ce7c275c ("event/dpaa2: add configuration functions")
Cc: stable@dpdk.org
Signed-off-by: Nipun Gupta <nipun.gupta@nxp.com>

eventdev: add new software timer adapter

This patch introduces a new version of the event timer adapter software
PMD. In the original design, timer event producer lcores in the primary
and secondary processes enqueued event timers into a ring, and a
service core in the primary process dequeued them and processed them
further.  To improve performance, this version does away with the ring
and lets lcores insert timers directly into timer skiplist data
structures; the service core directly accesses the lists as well, when
looking for timers that have expired.

To compare the burst and non-burst performance of the original and new
versions of the software event timer adapter, I ran the following
commands:

$ sudo ./build/app/dpdk-test-eventdev -c 0xFFE -s 0xC --vdev=event_sw0 \
-- --test=perf_queue --plcores=4,5,6 --wlcore=7,8,9 --stlist=p \
--prod_type_timerdev --worker_deq_depth=32

$ sudo ./build/app/dpdk-test-eventdev -c 0xFFE -s 0xC --vdev=event_sw0 \
-- --test=perf_queue --plcores=4,5,6 --wlcore=7,8,9 --stlist=p \
--prod_type_timerdev_burst --worker_deq_depth=32

With the new version, I see a 151% improvement in throughput for the
non-burst case, and a 270% improvement in throughput for the burst case.
I also see a 53% improvement in arm latency in the non-burst case and a
65% improvement in arm latency in the burst case.

Note: To perform the test,  I commented out a check in the original
version that checks the adapter tick interval against a minimum value.

Signed-off-by: Erik Gabriel Carrillo <erik.g.carrillo@intel.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>

eventdev: optimize Rx adapter enqueue

Setup event when the Rx queue is added to the
adapter in place of generating the event when it is
being enqueued to the event device.

Signed-off-by: Nikhil Rao <nikhil.rao@intel.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>

eventdev: remove event copy in Rx adapter

Remove copy from temporary event array on the stack to the
enqueue buffer event array entry, instead initialize event in the
enqueue buffer event array entry.

Suggested-by: Mattias Rönnblom <mattias.ronnblom@ericsson.com>
Signed-off-by: Nikhil Rao <nikhil.rao@intel.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>

mk: fix -march for octeontx2 target

Correct the wrong -march=-mcpu=armv8.2-a+crc+crypto+lse for
octeontx2 target. Since rte_cc_has_argument drops invalid
CFLAG and -mcpu=octeontx2 picks up the correct optimization,
this typo is not noticed in performance testing.

Fixes: 01d184798731 ("config: add octeontx2 machine")
Cc: stable@dpdk.org
Signed-off-by: Jerin Jacob <jerinj@marvell.com>

drivers: add reasons for components being disabled

For each driver where we optionally disable it, add in the reason why it's
being disabled, so the user knows how to fix it.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Luca Boccassi <bluca@debian.org>

lib: add reasons for components being disabled

For each library where we optionally disable it, add in the reason why it's
being disabled, so the user knows how to fix it.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Luca Boccassi <bluca@debian.org>

build: print list of disabled components

When configuring with meson we print out a list of enabled components, but
it is also useful to list out the disabled components and the reasons why.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Luca Boccassi <bluca@debian.org>

examples/vhost_crypto: support build with pkg-config

The vhost_crypto example app did not check for a libdpdk pkg-config file
and attempt to build using that. Add support for that method of compile to
align the app with the other examples.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Luca Boccassi <bluca@debian.org>

examples/vdpa: support build with pkg-config

The vdpa example app did not check for a libdpdk pkg-config file and
attempt to build using that. Add support for that method of compile to
align the app with the other examples.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Luca Boccassi <bluca@debian.org>

devtools: test pkg-config file

The pkg-config file generated as part of the build of DPDK should allow
applications to be built with an installed DPDK. We can test this as
part of the build by doing an install of DPDK to a temporary directory
within the build folder, and by then compiling up a few sample apps
using make working off that directory.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Luca Boccassi <bluca@debian.org>

devtools: test meson build with available compilers

Allow the script to run with a reduced set of builds if clang, or
other compilers, are missing.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Luca Boccassi <bluca@debian.org>

examples: support relocated DPDK install

For testing of DPDK, we want to override the prefix given by the
pkg-config file, so that we can get correct paths for DPDK installed
in an unusual location.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Luca Boccassi <bluca@debian.org>

rawdev: pass the device id as parameter to selftest

When running self-tests, the driver needs to know the device on which to
run the tests, so we need to take the device ID as parameter. Only the
skeleton driver is providing this selftest capability right now, so we can
easily update it for this change.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>

raw/skeleton: remove compile-time constant for device id

Use a variable value rather than compile-time constant zero as the
device id for the skeleton rawdev tests. This ensures we can make the
tests work even if other rawdevs are present.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>

test: add rawdev autotest to meson

the test_rawdev.c file was missing from the meson.build file, and the test
case from the list of test commands.

Fixes: 55ca1b0f2151 ("raw/skeleton: add test cases")
Cc: stable@dpdk.org
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>

buildtools: filter out checked symbols for icc

For some reason, ICC creates additional global symbols with a . which
triggers an error in the check that validates that symbols in the
experimental section are properly published in the map file of the
library. Filter them out.

Fixes: 3290ac14eb94 ("buildtools: detect discrepancies for experimental symbols")

Reported-by: Ferruh Yigit <ferruh.yigit@intel.com>
Signed-off-by: David Marchand <david.marchand@redhat.com>
Tested-by: Ferruh Yigit <ferruh.yigit@intel.com>

doc: remove useless Rx configuration in l2fwd guide

Update doc the match with code.

Fixes: 81f7ecd9 ("examples: use factorized default Rx/Tx configuration")
Cc: stable@dpdk.org
Signed-off-by: Bao-Long Tran <longtb5@viettel.com.vn>

examples/l3fwd-vf: remove unused Rx/Tx configuration

The RX and TX Prefetch, Host, and Write-back threshold values are
defined but not used anywhere. They are leftovers from a previous
patch.

Fixes: 81f7ecd9 ("examples: use factorized default Rx/Tx configuration")
Cc: stable@dpdk.org
Signed-off-by: Bao-Long Tran <longtb5@viettel.com.vn>

raw/skeleton: fix test of attribute set/get

Using memory on the stack instead of on the heap by calling malloc
also avoid memory leak in case of test case failures

Fixes: 88d0e47880ec ("raw/skeleton: fix memory leak on test failure")
Cc: stable@dpdk.org
Signed-off-by: Thinh Tran <thinhtr@linux.vnet.ibm.com>
Acked-by: Shreyansh Jain <shreyansh.jain@nxp.com>

malloc: deprecate unused function to set limit

The function rte_malloc_set_limit was defined but never implemented.
Mark it as deprecated for now, and remove in next release.

There is no point in keeping dead code.
"You Aren't Going to Need It"

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Anatoly Burakov <anatoly.burakov@intel.com>

eal/freebsd: add config reattach in secondary process

Linux EAL will attach the shared config at an arbitrary address,
find out where the shared config is mapped in the primary, and
then will reattach it at that exact address.

FreeBSD version doesn't seem to go for that extra reattach step,
which makes one wonder how did it ever work in the first place.

Fix the FreeBSD init to also reattach shared config to the exact
same place the primary process has it.

Fixes: 764bf26873b9 ("add FreeBSD support")
Cc: stable@dpdk.org
Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
Reviewed-by: David Marchand <david.marchand@redhat.com>

eal/freebsd: fix init completion

When init is complete, EAL is supposed to update internal config
to indicate that initialization is complete. Add missing write.

Fixes: a99c96e96ad3 ("eal: add internal flag of init completed")
Cc: stable@dpdk.org
Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
Reviewed-by: David Marchand <david.marchand@redhat.com>

mbuf: clarify outer offsets for non-tunnel packets

The default policy for offload-specific fields is that
they are undefined unless the corresponding offloads are
requested in mbuf ol_flags. This is also the case for outer
L2 and L3 length fields which must not be assumed to contain
zeros for non-tunnel packets. The patch clarifies this behaviour
in the comments and also adds appropriate checks to the PMDs which
do not check any tunnel-related offloads before using the said fields.

Signed-off-by: Ivan Malov <ivan.malov@oktetlabs.ru>
Reviewed-by: Andrew Rybchenko <arybchenko@solarflare.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>

examples/flow_filtering: remove out-of-date comment

Now we've setup both rx and tx queues.

Signed-off-by: Xiaolong Ye <xiaolong.ye@intel.com>
Acked-by: Ori Kam <orika@mellanox.com>

net/memif: support multi-process

Multi-process support for memif PMD.
Primary process handles connection establishment.
Secondary process queries for memory regions.

Signed-off-by: Jakub Grajciar <jgrajcia@cisco.com>
Acked-by: Anatoly Burakov <anatoly.burakov@intel.com>

net/af_xdp: support need wakeup feature

This patch enables need_wakeup flag for Tx and fill rings, when this
flag is set by the driver, it means that the userspace application has
to explicitly wake up the kernel Rx or kernel Tx processing by issuing
a syscall. Poll() can wake up both and sendto() or its alternatives
will wake up Tx processing only.

This feature is to provide efficient support for case that application
and driver executing on the same core.

Signed-off-by: Xiaolong Ye <xiaolong.ye@intel.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net: fix how L4 checksum choice is tested

The API to prepare checksum offloads mistreats L4
checksum type enum values as self-contained flags.

Turning these flag checks into enum checks causes
warnings by GCC about possibly uninitialised IPv4
header pointer. The issue was found to show up in
the case of GCC versions 4.8.5 and 5.4.0, however,
it might be the case for a wider variety of other
versions. Initialise the pointer upon declaration.
and explain the reason behind this in the comment.

Fixes: 4fb7e803eb1a ("ethdev: add Tx preparation")
Cc: stable@dpdk.org
Signed-off-by: Ivan Malov <ivan.malov@oktetlabs.ru>
Reviewed-by: Andrew Rybchenko <arybchenko@solarflare.com>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>

maintainers: update for szedata2 PMD

Rastislav will co-maintain the szedata2 PMD with me.

Signed-off-by: Jan Remes <remes@netcope.com>

net/pcap: enable infinitely Rx a pcap file

It can be useful to use pcap files for some rudimental performance
testing. This patch enables this functionality in the pcap driver.

At a high level, this works by creating a ring of sufficient size to
store the packets in the pcap file passed to the application. When the
rx function for this mode is called, packets are dequeued from the ring
for use by the application and also enqueued back on to the ring to be
"received" again.

A tx_drop mode is also added since transmitting to a tx_pcap file isn't
desirable at a high traffic rate.

Jumbo frames are not supported in this mode. When filling the ring at rx
queue setup time, the presence of multi segment mbufs is checked for.
The PMD will exit on detection of these multi segment mbufs.

Signed-off-by: Cian Ferriter <cian.ferriter@intel.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/pcap: use a struct to pass user options

The argument lists on some of the device creation functions are quite
large. Using a struct to hold the user options parsed in
'pmd_pcap_probe' will allow for cleaner function calls and definitions.
Adding user options will also be easier.

Signed-off-by: Cian Ferriter <cian.ferriter@intel.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>

doc: fix typos in flow API guide

Fixes: 3e0ceb9f17ff ("doc: add basic howto for flow API")
Cc: stable@dpdk.org
Signed-off-by: Xiaolong Ye <xiaolong.ye@intel.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/netvsc: fix xstats for VF device

The id values for VF stats were not being offset correctly.
And getting xstats for VF device only worked if VF device supported
it; it did not support the generic stats.

Fixes: dc7680e8597c ("net/netvsc: support integrated VF")
Cc: stable@dpdk.org
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>

net/netvsc: fix xstats id

The xstats_get was not setting id correctly for each entry.

Fixes: 4e9c73e96e83 ("net/netvsc: add Hyper-V network device")
Cc: stable@dpdk.org
Signed-off-by: Mohsin Mazhar Shaikh <mohsinmazhar_shaikh@trendmicro.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>

net/netvsc: support configuring RSS parameters

Add RSS hash key and reta update and query functions.

Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>

net/netvsc: fix RSS offload settings

Use the result from querying host to show the RSS info.

Fixes: 4e9c73e96e83 ("net/netvsc: add Hyper-V network device")
Cc: stable@dpdk.org
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>

net/mlx5: fix 32-bit build

This is to fix the error:
  ```
  drivers/net/mlx5/mlx5_defs.h:14:26:
  error: format '%lx' expects argument of type 'long unsigned int',
  but argument 5 has type 'off_t {aka long long int}' [-Werror=format=]
  drivers/net/mlx5/mlx5_txq.c:569:48: note: format string is defined here
     DRV_LOG(DEBUG, "port %u: uar_mmap_offset 0x%lx"
                                                ~~^
                                                %llx
  ```
Which reproduces with gcc (Ubuntu 7.4.0-1ubuntu1~18.04.1) 7.4.0.

Fixes: 6bf10ab69be0 ("net/mlx5: support 32-bit systems")
Cc: stable@dpdk.org
Signed-off-by: Ali Alnubani <alialnu@mellanox.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net: fix encapsulation markers for inner L3 offset

The API to prepare checksum offloads employs outer
IP checksum flag to tell regular IPv4 packets from
tunnel packets with outer IPv4 encapsulation. This
flag cannot serve as a marker for the said purpose
because a packet can have outer IPv4 encapsulation
and may not have outer IP checksum offload request.

Fix the API by changing the said criterion to test
outer IPv4 flag rather than outer IP checksum flag.
Use simpler spelling of the conditional expression.

Fixes: 4fb7e803eb1a ("ethdev: add Tx preparation")
Cc: stable@dpdk.org
Signed-off-by: Ivan Malov <ivan.malov@oktetlabs.ru>
Reviewed-by: Andrew Rybchenko <arybchenko@solarflare.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>

net/hinic: add RSS stats and promiscuous ops

Add RSS, stats, promiscuous ops related function codes.

Signed-off-by: Ziyang Xuan <xuanziyang2@huawei.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/hinic: add Rx/Tx

This patch add package sending and receiving function codes.

Signed-off-by: Ziyang Xuan <xuanziyang2@huawei.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/hinic: add start/stop and queue ops

Add dev_start, dev_stop, link_update, queue_setup, queue_release
related function codes.

Signed-off-by: Ziyang Xuan <xuanziyang2@huawei.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/hinic: add device initialization

Add device initialization function codes.

Signed-off-by: Ziyang Xuan <xuanziyang2@huawei.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/hinic: add build and doc files

Add build and doc files along with hinic_pmd_ethdev.c
which just includes PMD register and log initialization
for compilation.

Signed-off-by: Ziyang Xuan <xuanziyang2@huawei.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/hinic/base: add various headers

Add various headers that define mgmt commands, cmdq commands
and basic defines for use in the code.

Signed-off-by: Ziyang Xuan <xuanziyang2@huawei.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/hinic/base: support context and work queue

Work queue is used for cmdq and Rx/Tx buff description.
NIC business needs to configure cmdq context and txq/rxq
context. This patch adds data structures and function codes
for work queue and context.

Signed-off-by: Ziyang Xuan <xuanziyang2@huawei.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/hinic/base: add NIC business configurations

The items of configurations and queries for NIC business include
MAC, VLAN, MTU, RSS and so on. These configurations and queries
are handled by mgmt module. This patch introduces related
data structures and function codes.

Signed-off-by: Ziyang Xuan <xuanziyang2@huawei.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/hinic/base: add hardware operation

Add code for hardware operation, including configuration,
query and so on.

Signed-off-by: Ziyang Xuan <xuanziyang2@huawei.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/hinic/base: add mgmt module

Mgmt module is a kind of administration module for the chip.
It is responsible for handling administration command from host.
It uses api channel. This patch adds related data structures,
packaged interfaces and function codes.

Signed-off-by: Ziyang Xuan <xuanziyang2@huawei.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/hinic/base: add eq mechanism

Eqs include aeq and ceq. PMD supports aeq only. Aeq is a kind of
queue for mgmt asynchronous message and mgmt command response message.
This patch introduces data structures, initialization, and related
interfaces about aeq.

Signed-off-by: Ziyang Xuan <xuanziyang2@huawei.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/hinic/base: support cmdq mechanism

Microcode is named ucode in Hi1822. Its main responsibility is data
transmission and reception. But it can also handle some administration
works. It uses cmdq mechanism. This patch introduces data structures,
initialization, interfaces, and commands sending functions of cmdq.

Signed-off-by: Ziyang Xuan <xuanziyang2@huawei.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/hinic/base: add API command channel

API command channel is for communicating with mgmt module of chip.
This patch introduces data structures, initialization, interfaces,
and commands sending functions of API command channel.

Signed-off-by: Ziyang Xuan <xuanziyang2@huawei.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/hinic/base: add HW interfaces of BAR operation

This patch adds some HW interfaces for bar operation interfaces,
including: mapped bar address getting, HW attributes getting,
msi-x reg operation, function type getting and so on.

Signed-off-by: Ziyang Xuan <xuanziyang2@huawei.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/hinic/base: add HW registers definition

Add HW registers definition header file for Hi1822 NIC.

Signed-off-by: Ziyang Xuan <xuanziyang2@huawei.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

drivers/net: fix double free on init failure

If we don't clear mac_addrs, ethdev will double free it on cleanup.

Fixes: e16adf08e54d ("ethdev: free all common data when releasing port")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/bnx2x: fix invalid free on unplug

mac_addrs points to a field in dev_private.
We can't let ethdev free it.

Fixes: e16adf08e54d ("ethdev: free all common data when releasing port")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/nfb: avoid memory leak on unplug

Clearing mac_addrs on remove will prevent ethdev from freeing it.

Fixes: 6435f9a0ac22 ("net/nfb: add new netcope driver")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/avp: remove resources when port is closed

The rte_eth_dev_close() function now handles freeing resources for
devices (e.g., mac_addrs).  To conform with the new close() behaviour we
are asserting the RTE_ETH_DEV_CLOSE_REMOVE flag so that
rte_eth_dev_close() releases all device level dynamic memory.

Second level memory allocated to each individual rx/tx queue is now
freed as part of the close() operation therefore making it safe for the
rte_eth_dev_close() function to free the device private data without
orphaning the rx/tx queue pointers.

Signed-off-by: Allain Legacy <allain.legacy@windriver.com>
Acked-by: Matt Peters <matt.peters@windriver.com>

net/softnic: remove trace of legacy linuxapp

This check on Linux environment has been added at a time when we already
had switched to using the boolean RTE_EXEC_ENV_LINUXAPP.
It was then missed when converting to RTE_EXEC_ENV_LINUX.

Fixes: 7709a63bf178 ("net/softnic: add connection agent")
Fixes: 742bde12f3bd ("build/linux: rename macro from LINUXAPP to LINUX")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/ice/base: update Tx context struct

Add internal usage flag, bit 91 as described in spec.
Update width of internal queue state to 122 also as described in spec.

Signed-off-by: Ashish Shah <ashish.n.shah@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: change flow and profile removal

- Add function to remove ES profile map as it is now being used in
clearing and freeing HW tables.
- Locks were initially not used for releasing ES profile maps and
flow profiles as the sequence is part of driver unload. Adding
calls to acquire and release locks to ensure that any calls made
by the VF VSI during VFR or unload do not result in memory
access violations.

Signed-off-by: Vignesh Sridhar <vignesh.sridhar@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: update driver unloading field

According to recent specification versions, the field in the Queue
Shutdown AdminQ command consisting of the "driver unloading" indication
is not a 4 byte field (it is byte.bit 16.0). Change it to a byte and
remove the unnecessary endian conversion.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: reduce calls to get profile associations

Added refresh_required flag to determine if we need to update the
recipe to profile mapping cache. This will reduce the number of
calls made to refresh the profile map.

Signed-off-by: Shivanshu Shukla <shivanshu.shukla@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: fix NVGRE header structure

Correct NVGRE header structure and its field offsets.

Fixes: 04b8ec1ea807 ("net/ice/base: add protocol structures and defines")
Cc: stable@dpdk.org
Signed-off-by: Dan Nowlin <dan.nowlin@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: refactor zeroing bitmap size

A couple of places in the code use a 'sizeof(bitmap) * BITS_PER_BYTE'
construction to calculate the size of the bitmap when calling
ice_zero_bitmap. Instead of doing this, just use the same value as in
the ice_declare_bitmap declaration.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: refactor zeroing bitmap

A few places in the code used ice_memset instead of ice_zero_bitmap to
initialize a bitmap to zeros.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: fix copying bitmap

Fix ice_cp_bitmap to copy the entire storage.

Fixes: c9e37832c95f ("net/ice/base: rework on bit ops")
Cc: stable@dpdk.org
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: fix zeroing bitmap

Fix ice_zero_bitmap to zero the entire storage.

Fixes: c9e37832c95f ("net/ice/base: rework on bit ops")
Cc: stable@dpdk.org
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: deduce TSA value in the CEE mode

In CEE mode, the TSA information can be derived from the reported
priority value.

Signed-off-by: Chinh T Cao <chinh.t.cao@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: reduce adding advanced rule time

While calling ice_find_recp we were calling ice_get_recp_to_prof_map
every time we called ice_find_recp. ice_get_recp_to_prof_map is a very
expensive operation and we should try to reduce the number of times we
call this function. So moved it into ice_get_recp_frm_fw since we only
need to have fresh recp_to_profile mapping when we we check FW to see if
the recipe we are trying to add already exists in FW.

Signed-off-by: Shivanshu Shukla <shivanshu.shukla@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: allow forward to Q groups in switch rule

Enable forward to Q group action in ice_add_adv_rule.

Signed-off-by: Dan Nowlin <dan.nowlin@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: enable additional switch rules

Add capability to create inner IP and inner TCP switch recipes and
rules. Change UDP tunnel dummy packet to accommodate the training of
these new rules.

Signed-off-by: Dan Nowlin <dan.nowlin@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: add VXLAN/generic tunnel management

Added routines for handling tunnel management:
- ice_tunnel_port_in_use()
- ice_tunnel_get_type()
- ice_find_free_tunnel_entry()
- ice_create_tunnel()
- ice_destroy_tunnel()

Signed-off-by: Dan Nowlin <dan.nowlin@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: add some minor new defines

1. Add defines for Link Topology Netlist Section.
2. Add missing Read MAC command response bits.
3. Adds AQ error 29.

Signed-off-by: Jacek Naczyk <jacek.naczyk@intel.com>
Signed-off-by: Lev Faerman <lev.faerman@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: refactor VSI node scheduler

Refactored VSI node sched code to use port_info ptr as call arg.

The declaration of VSI node getter function has been modified to use
pointer to ice_port_info structure instead of pointer to hw structure.
This way suitable port_info structure is used to find VSI node.

Signed-off-by: Grzegorz Nitka <grzegorz.nitka@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: cleanup flex pipe files

Make functions that can be, static. Remove some code that is not
currently called.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: cleanup unused code

Cleanup the useless code.

Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: update marker to mark unused func parameters

use __ALWAYS_UNUSED to mark function parameters unused to replace
__always_unused marker.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: introduce some new macros

Introduce some more new macros, like ICE_VSI_LB and the like.

Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: fix Rx functionality for ethertype filters

In the function ice_add_eth_mac(), there is a line that
hard-codes the filter info flag to TX. This is redundant
and inaccurate. That flag will be set by the calling
function that built the list of filters to add, and
hard-coding it eliminates the Rx functionality of this
code. The paired function ice_remove_eth_mac() does not
do this, making a mis-matched pair.

Fixes: 157d00901f97 ("net/ice/base: add functions for ethertype filter")
Cc: stable@dpdk.org
Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: protect list add with lock

Function ice_add_rule_internal needs to call ice_create_pkt_fwd_rule
with lock held because it uses the LIST_ADD to modify the filter
rule list. It needs to be protected when modified.

Signed-off-by: Tarun Singh <tarun.k.singh@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: check RSS flow profile list

Minor change to check if there are any RSS flow profiles to remove.
This will avoid flow profile lock acquisition and release
if the list is empty.

Signed-off-by: Vignesh Sridhar <vignesh.sridhar@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: add function to deallocate flow entry

Decouple ice_dealloc_flow_entry from ice_flow_rem_entry_sync.

Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/ice/base: fix signed package download

In order to properly support signed packages, we always have
to send the complete buffer to firmware, regardless of any
unused space at the end. This is because the SHA hash value
is computed over the entire buffer.

Fixes: 51d04e4933e3 ("net/ice/base: add flexible pipeline module")
Cc: stable@dpdk.org
Signed-off-by: Dan Nowlin <dan.nowlin@intel.com>
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Signed-off-by: Leyi Rong <leyi.rong@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>