igb_uio: fix unexpected removal for hot-unplug

When a device is hot-unplugged, pci_remove will be invoked unexpectedly
before pci_release, it will caused kernel hung issue which will throw the
error info of "Trying to free already-free IRQ XXX". And on the other hand,
if pci_remove before pci_release, the interrupt will not got chance to be
disabled. So this patch aim to fix this issue by adding pci_release call
in pci_remove, it will gurranty that all pci clean up will be done before
pci removal.

Signed-off-by: Jeff Guo <jia.guo@intel.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>

raw/skeleton: fix memory leak on test failure

In skeleton_rawdev unit tests, a malloc'd memory was leaking in case
the next sequential test fails. This fix moves the free of the
malloc'd memory above the failing test.

Coverity issue: 260402
Fixes: 55ca1b0f2151 ("raw/skeleton: add test cases")

Signed-off-by: Shreyansh Jain <shreyansh.jain@nxp.com>

net/dpaa2: decrease link state log level

In case the link is down during initial link state check, messages for
link state check flood the console. Reducing the log level for these.

Signed-off-by: Shreyansh Jain <shreyansh.jain@nxp.com>

bus/fslmc: ignore dpaax PA-VA table errors

Presence of PA-VA Table is transparent to the drivers. Ignoring the
return values from table update call.

Signed-off-by: Shreyansh Jain <shreyansh.jain@nxp.com>
Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>

common/dpaax: reduce log level

DPAAX is a library used by various NXP drivers. In case of non-NXP
environment, this start spewing message about unavailability of
necessary environment.

This patch reduces the log level for certain messages as well as
reduces overall log-level. As a library, these message are not
necessarily relevant at higher log level, either.

Signed-off-by: Shreyansh Jain <shreyansh.jain@nxp.com>
Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Tested-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>

common/dpaax: fix nodes check

In case the memory for nodes cannot be allocated, there is no need
to check for the length. Also, `node_count` is an unsigned value
and cannot be less than 0.

Coverity issue: 323521
Fixes: 2f3d633aa593 ("common/dpaax: add library for PA/VA translation table")

Signed-off-by: Shreyansh Jain <shreyansh.jain@nxp.com>

common/dpaax: fix uninitialized PA-VA table case

There is a possibility that either because of missing device tree entry
or lack of memory, the PA-VA table might not be available. But, the
table being transparent, the callers don't necessary check for its
initialization state. This is explicitly done during update and
translation call.

Fixes: 2f3d633aa593 ("common/dpaax: add library for PA/VA translation table")

Signed-off-by: Shreyansh Jain <shreyansh.jain@nxp.com>

mk: use EXTRA_CFLAGS for pmdinfogen

Currently not able to pass EXTRA_CFLAGS while building *.pmd.c file,
adding it.

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>

crypto/dpaa2_sec: support PDCP offload

PDCP session configuration for lookaside protocol offload
and data path is added.

Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Signed-off-by: Akhil Goyal <akhil.goyal@nxp.com>

crypto/dpaa2_sec: add sample PDCP descriptor APIs

DPAA2 SEC platform can support look aside protocol
offload for PDCP protocol.

The relevant APIs for configuring the hardware for PDCP
is added for various modes and crypto algorithms.

Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Signed-off-by: Horia Geanta Neag <horia.geanta@nxp.com>
Signed-off-by: Alex Porosanu <alexandru.porosanu@nxp.com>
Signed-off-by: Akhil Goyal <akhil.goyal@nxp.com>

security: support PDCP

Packet Data Convergence Protocol (PDCP) is added in rte_security
for 3GPP TS 36.323 for LTE.

The patchset provide the structure definitions for configuring the
PDCP sessions and relevant documentation is added.

Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Signed-off-by: Akhil Goyal <akhil.goyal@nxp.com>
Acked-by: Anoob Joseph <anoob.joseph@caviumnetworks.com>

crypto/caam_jr: fix type redefinition

dma_addr_t is already defined in compat.h.
so removing the local definition from caam_jr_config.h

Fixes: 64c0451f5bb9 ("crypto/caam_jr: add HW tuning options")

Signed-off-by: Gagandeep Singh <g.singh@nxp.com>

drivers: fix build if security lib disabled

RTE_SECURITY is enabled by default. If it is disabled, dpaa2_sec,
dpaa_sec and caam_jr compilation fails.

This patch fixes compilation by disabling these drivers
when rte_security is not available.

Fixes: 1ee9569576f6 ("config: enable dpaaX drivers for generic ARMv8")
Fixes: 09e1e8d256b0 ("mk: fix dependencies of dpaaX drivers")
Fixes: af7c9b5e9ce7 ("crypto/caam_jr: introduce basic driver")
Cc: stable@dpdk.org
Signed-off-by: Akhil Goyal <akhil.goyal@nxp.com>

ipc: fix undefined behavior in no-shconf mode

In no-shconf mode the rte_mp_request_sync() wasn't initializing
the `reply` parameter, which contained e.g. a number of sent
requests. Callers of rte_mp_request_sync() might check that
param afterwards and might read potentially unitialized memory.

The no-shconf check that makes us return early (with rc = 0) was
placed before the `reply` initialization. Fix this by making the
`reply` initialization occur first.

Fixes: 5848e3d2813c ("ipc: support --no-shconf mode")
Cc: stable@dpdk.org
Signed-off-by: Dariusz Stojaczyk <dariuszx.stojaczyk@intel.com>
Acked-by: Anatoly Burakov <anatoly.burakov@intel.com>

kvargs: fix processing a null list

In the doxygen description of rte_kvargs_process(), it is said:
If *kvlist* is NULL function does nothing.
It has been added by mistake here instead of rte_kvargs_free().
Anyway, null list should be correctly handled in both functions.

Comments are fixed in both functions and NULL handling is added
to rte_kvargs_process().

Fixes: c34af7424e09 ("kvargs: fix freeing behaviour for null")
Cc: stable@dpdk.org
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Olivier Matz <olivier.matz@6wind.com>

mem: fix resource leak

Segment preallocation code allocates an array of structures on the
heap but does not free the memory afterwards. Fix it by freeing it
at the end of the function, and changing control flow to always go
through that code path.

Coverity issue: 323524
Fixes: 1dd342d0fdc4 ("mem: improve segment list preallocation")
Cc: stable@dpdk.org
Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>

test: fix build of external memory test

There was a compilation error in test_external_mem.c:

  CC test_external_mem.o
test_external_mem.c: In function ‘test_external_mem’:
test_external_mem.c:375:2: error: ‘for’ loop initial declarations are
                           only allowed in C99 mode

  for (int i = 0; i < n_pages; i++) {
  ^
test_external_mem.c:375:2: note: use option -std=c99 or -std=gnu99 to
                           compile your code

Fixes: b270daa43b3d ("test: support external memory")

Signed-off-by: Dan Gora <dg@adax.com>
Acked-by: Anatoly Burakov <anatoly.burakov@intel.com>

eal: fix bus name read for removal in multi-process

A crash may appear when removing some PCI devices because
dev->devargs is not always initialized. So use dev->bus instead of
dev->devargs->bus when building devargs string to remove a device.

Fixes: 244d5130719c ("eal: enable hotplug on multi-process")

Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>

bus/vdev: fix uninitialized device bus

Device bus should be initialized after bus scan.
While it does not happened when scan vdev from secondary process,
that cause segment fault at rte_dev_probe when call dev->bus->xxx.

Fixes: cdb068f031c6 ("bus/vdev: scan by multi-process channel")
Cc: stable@dpdk.org
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>

mem: improve segment list preallocation

Current code to preallocate segment lists is trying to do
everything in one go, and thus ends up being convoluted,
hard to understand, and, most importantly, does not scale beyond
initial assumptions about number of NUMA nodes and number of
page sizes, and therefore has issues on some configurations.

Instead of fixing these issues in the existing code, simply
rewrite it to be slightly less clever but much more logical, and
provide ample comments to explain exactly what is going on.

We cannot use the same approach for 32-bit code because the
limitations of the target dictate current socket-centric
approach rather than type-centric approach we use on 64-bit
target, so 32-bit code is left unmodified. FreeBSD doesn't
support NUMA so there's no complexity involved there, and thus
its code is much more readable and not worth changing.

Fixes: 1d406458db47 ("mem: make segment preallocation OS-specific")
Cc: stable@dpdk.org
Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>

eal: improve musl compatibility of thread log

Musl complains about pthread id being of wrong size, because on
musl, pthread_t is a struct pointer, not an unsigned int. Fix the
printing code by casting pthread id to unsigned pointer type and
adjusting the format specifier to be of appropriate size.

Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>

eal: improve musl compatibility of string functions

Musl wraps various string functions such as strlcpy in order to
harden them. However, the fortify wrappers are included without
including the actual string functions being wrapped, which
throws missing definition compile errors. Fix by including
string.h in string functions header.

Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>

mem: improve musl compatibility

When built against musl, fcntl.h doesn't silently get included.
Fix by including it explicitly.

Bugzilla ID: 31

Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>

eal/linux: improve musl compatibility

When built against musl, fcntl.h doesn't silently get included.
Fix by including it explicitly.

Bugzilla ID: 33

Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>

fbarray: improve musl compatibility

When built against musl, fcntl.h doesn't silently get included.
Fix by including it explicitly.

Bugzilla ID: 34

Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>

vfio: improve musl compatibility

Musl already has PAGE_SIZE defined, and our define clashed with it.
Rename our define to SYS_PAGE_SIZE.

Bugzilla ID: 36

Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>

mk: build with _GNU_SOURCE defined by default

We use _GNU_SOURCE all over the place, but often times we miss
defining it, resulting in broken builds on musl. Rather than
fixing every library's and driver's and application's makefile,
fix it by simply defining _GNU_SOURCE by default for all
builds.

Remove all usages of _GNU_SOURCE in source files and makefiles,
and also fixup a couple of instances of using __USE_GNU instead
of _GNU_SOURCE.

Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>

devargs: fix freeing during device removal

After calling unplug function of a bus, the device is expected
to be freed. It is too late for getting devargs to remove.
Anyway, the buses which implement unplug are already freeing
the devargs, except the PCI bus.
So the call to rte_devargs_remove() is removed from EAL and
added in PCI.

Fixes: 2effa126fbd8 ("devargs: simplify parameters of removal function")

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>

app/testpmd: set packet dump based on verbosity level

when changing verbosity level it will configure rx/tx callbacks to dump
packets based on the verbosity value as following:
    1- dump only received packets:
       testpmd> set verbose 1
    2- dump only sent packets:
       testpmd> set verbose 2
    3- dump sent and received packets:
       testpmd> set verbose (any number > 2)
    4- disable dump
       testpmd> set verbose 0

Signed-off-by: Raslan Darawsheh <rasland@mellanox.com>
Acked-by: Bernard Iremonger <bernard.iremonger@intel.com>

app/testpmd: add packet dump callbacks

add new rx/tx callback functions to be used for dumping the packets.

Signed-off-by: Raslan Darawsheh <rasland@mellanox.com>
Acked-by: Bernard Iremonger <bernard.iremonger@intel.com>

app/testpmd: move dumping packets to a separate function

verbosity for the received/sent packets is needed in all of the
forwarding engines so moving it to be in a separate function

Acked-by: Bernard Iremonger <bernard.iremonger@intel.com>
Signed-off-by: Raslan Darawsheh <rasland@mellanox.com>

app/testpmd: fix physical port socket initialization

Once the lcore list setting excluded the socket which physical device
attached, it will cause failure. Meanwhile, it will disable Testpmd
cross NUMA scenario.

Fixes: dbfb8ec7094c ("app/testpmd: optimize mbuf pool allocation")
Cc: stable@dpdk.org
Signed-off-by: Phil Yang <phil.yang@arm.com>
Reviewed-by: Gavin Hu <gavin.hu@arm.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>

net/tap: fix reported number of Tx packets

When writev fails to send packets it doesn't update the
number of Tx packets, but it still num_tx is updated.

The value that should be returned is the actual number
of sent packets which is num_packets.

Fixes: 02f96a0a82d1 ("net/tap: add TUN/TAP device PMD")
CC: stable@dpdk.org
Signed-off-by: Raslan Darawsheh <rasland@mellanox.com>
Acked-by: Keith Wiles <keith.wiles@intel.com>

net/ixgbe: remove mild profanity

At the tail end of comment about barriers (I feel your pain);
remove mild profanity.

Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/bnx2x: remove profanity

No need for profanity in comments.

Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

doc: clarify VLAN and QinQ Tx offload prerequisite

- Fix missing PKT_TX_VLAN mbuf.ol_flag and mbuf.vlan_tci
fields for Tx VLAN INSERT offload.

- Fix missing mbuf.vlan_tci_outer field for Tx QINQ INSERT offload.

- Rename deprecated PKT_TX_QINQ_PKT to PKT_TX_QINQ

Fixes: cba7f53b717d ("ethdev: introduce Tx queue offloads API")
Cc: stable@dpdk.org
Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>

mbuf: fix missing Tx outer UDP checksum flag name

Fix missing Tx outer udp checksum flag name

Fixes: df694a05bfff ("ethdev: add Tx offload outer UDP checksum definition")

Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>

mbuf: fix offload flag name and list

Fix missing PKT_TX* & PKT_RX* ol_flag name and fix ol_flag list.

Fixes: 6d18505efaa6 ("vhost: support UDP Fragmentation Offload")
Fixes: 829a1c2c41dc ("mbuf: extend flow director field")
Fixes: 63c0d74daaa9 ("mbuf: add Tx side tunneling type")
Cc: stable@dpdk.org
Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/mlx5: fix build on Arm

On some ARM environment, the below compilation error will be seen

dpdk/drivers/net/mlx5/mlx5_flow_dv.c: In function
'flow_dv_translate_item_nvgre':
/tmp/dpdk/drivers/net/mlx5/mlx5_flow_dv.c:785:22: error: pointer targets
in initialization differ in signedness [-Werror=pointer-sign]
  const char *tni_v = nvgre_v->tni;

The reason for this error is that nvgre_v->tni is defined as byte array
in size of 3B. However the code in the function iterate till the 4B in
order to copy/set also the subsequent field after it (flow_id)

Fixing by pointing to this struct from a different pointer.

Fixes: fc2c498ccb94 ("net/mlx5: add Direct Verbs translate items")

Signed-off-by: Shahaf Shuler <shahafs@mellanox.com>
Acked-by: Ori Kam <orika@mellanox.com>

vhost: enable postcopy protocol feature

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>

net/vhost: add parameter to enable postcopy

Introduce a new postcopy-support parameter to Vhost PMD that
passes the RTE_VHOST_USER_POSTCOPY_SUPPORT flag at vhost
device register time.

Flag should only be set if application does not prefault guest
memory using, for example, mlockall() syscall.

Default value is 0, meaning that postcopy support is disabled
unless specified explicitly.

Example to enable postcopy support for a given device:

--vdev 'net_vhost0,iface=/tmp/vhost-user1,postcopy-support=1'

Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>

vhost: restrict postcopy live-migration enablement

Postcopy live-migration feature requires the application to
not populate the guest memory. As the vhost library cannot
prevent the application to that (e.g. preventing the
application to call mlockall()), the feature is disabled by
default.

The application should only enable the feature if it does not
force the guest memory to be populated.

In case the user passes the RTE_VHOST_USER_POSTCOPY_SUPPORT
flag at registration but the feature was not compiled,
registration fails.

For the same reason, postcopy and dequeue zero copy features
are not compatible, so don't advertize postcopy support if
dequeue zero copy is requested.

Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>

vhost: support postcopy end request

The master sends this message before stopping handling
userfaults, so that the backend closes the userfaultfd.

The master waits for the slave to acknowledge the request
with an empty 64bits payload for synchronization purpose.

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>

vhost: send userfault range addresses back to Qemu

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>

vhost: avoid useless VhostUserMemory copy

The VHOST_USER_SET_MEM_TABLE payload is copied when handled,
whereas it could directly be referenced.

This is not very important, but next, we'll need to update the
payload and send it back to Qemu.

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Acked-by: Ilya Maximets <i.maximets@samsung.com>

vhost: register new regions with userfaultfd

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>

vhost: support postcopy listen message

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>

vhost: introduce postcopy advise message

This patch opens a userfaultfd and sends it back to Qemu's
VHOST_USER_POSTCOPY_ADVISE request.

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>

vhost: add config flag for postcopy

Postcopy live-migration features relies on userfaultfd,
which was only introduced in kernel v4.3.

This patch introduces a new define to allow building vhost
library on kernels not supporting userfaultfd.

With legacy build system, user has to explicitly set
CONFIG_RTE_LIBRTE_VHOST_POSTCOPY to 'y'.

With Meson build system, RTE_LIBRTE_VHOST_POSTCOPY gets
automatically defined if userfaultfd kernel header is
present.

Suggested-by: Ilya Maximets <i.maximets@samsung.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Acked-by: Ilya Maximets <i.maximets@samsung.com>

vhost: enable fds passing in vhost-user messages

Passing userfault fds to Qemu will be required for postcopy
live-migration feature.

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>

vhost: pass socket fd to message handling callbacks

This is not used for now, but will be needed for the
special handling of VHOST_USER_SET_MEM_TABLE message
once postcopy will be supported.

Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>

vhost: add number of fds to vhost-user messages

As soon as some ancillary data (fds) are received, it is copied
without checking its length.

This patch adds the number of fds received to the message,
which is set in read_vhost_message().

This is preliminary work to support sending fds to Qemu.

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>

vhost: define postcopy protocol flag

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Acked-by: Ilya Maximets <i.maximets@samsung.com>
Reviewed-by: Tiwei Bie <tiwei.bie@intel.com>

vhost: fix error handling when mem table gets updated

When the memory table gets updated, the rings addresses need
to be translated again. If it fails, we need to exit cleanly
by unmapping memory regions.

Fixes: d5022533c20a ("vhost: retranslate vring addr when memory table changes")
Cc: stable@dpdk.org
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Acked-by: Ilya Maximets <i.maximets@samsung.com>

vhost: fix payload size of reply

QEMU doesn't expect any payload for the reply of
VHOST_USER_SET_LOG_BASE request, so don't send any.
Note that the Vhost-user specification isn't clear about
it and would need to be fixed.

Fixes: 54f9e32305d4 ("vhost: handle dirty pages logging request")
Cc: stable@dpdk.org
Reported-by: Ilya Maximets <i.maximets@samsung.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Acked-by: Ilya Maximets <i.maximets@samsung.com>

vhost: clarify reply-ack in case a reply was already sent

For messages that require a reply, a second ack should not be
sent when reply-ack protocol feature is negotiated, even if
the corresponding flag is set in the message.

The code is compliant with the spec but it isn't clear it is,
so this patch adds a comment to make it explicit.

Suggested-by: Ilya Maximets <i.maximets@samsung.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Acked-by: Ilya Maximets <i.maximets@samsung.com>
Reviewed-by: Tiwei Bie <tiwei.bie@intel.com>

vhost: fix return code of messages requiring replies

VHOST_USER_GET_PROTOCOL_FEATURES, VHOST_USER_GET_VRING_BASE
and VHOST_USER_SET_LOG_BASE require replies, so their handlers
should return VH_RESULT_REPLY, not VH_RESULT_OK.

Fixes: 0bff510b5ea6 ("vhost: unify message handling function signature")

Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Acked-by: Ilya Maximets <i.maximets@samsung.com>
Reviewed-by: Tiwei Bie <tiwei.bie@intel.com>

vhost: fix messages results handling

Return of message handling has now changed to an enum that can
take non-negative value that is not zero in case a reply is
needed. But the code checking the variable afterwards has not
been updated, leading to success messages handling being
treated as errors.

External post and pre callbacks return type needs also to be
changed to the new enum, so that its handling is consistent.
This is done in this patch alongside with the convertion of
its only user, vhost-crypto backend.

Fixes: 0bff510b5ea6 ("vhost: unify message handling function signature")

Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>

vhost: add doxygen comment to vDPA header

As APIs in rte_vdpa.h are public, we need to add doxygen comments
to all APIs and structures.

Signed-off-by: Xiaolong Ye <xiaolong.ye@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>

vhost: fix notification for packed ring

The notification can't be disabled in packed ring when
application tries to disable notification, because the
device event flags field is overwritten by an unexpected
value. This patch fixes this issue.

Fixes: b1cce26af1dc ("vhost: add notification for packed ring")
Cc: stable@dpdk.org
Signed-off-by: Tiwei Bie <tiwei.bie@intel.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>

net/mlx5: rewrite MAC address by E-Switch

Offload following modify MAC address actions to E-Switch
via TC-Flower driver

- RTE_FLOW_ACTION_TYPE_SET_MAC_SRC
- RTE_FLOW_ACTION_TYPE_SET_MAC_DST

The corresponding rte_flow_item_eth must be present in
rte_flow pattern

Only support modify outer layer MAC address

The example testpmd command is:

    flow create 0 transfer ingress
         pattern eth / ipv4 / udp dst is 7000 / end
         actions set_mac_dst mac_addr dd:00:aa:11:bb:33 /
         set_mac_src mac_addr bb:00:cc:11:aa:22 /
 port_id id 1 / end

Signed-off-by: Xiaoyu Min <jackmin@mellanox.com>
Acked-by: Yongseok Koh <yskoh@mellanox.com>

app/testpmd: add commands to modify MAC address

add commands to support following actions:

- RTE_FLOW_ACTION_TYPE_SET_MAC_SRC
- RTE_FLOW_ACTION_TYPE_SET_MAC_DST

Signed-off-by: Xiaoyu Min <jackmin@mellanox.com>
Acked-by: Yongseok Koh <yskoh@mellanox.com>

ethdev: add generic MAC address rewrite actions

rte_flow actions:
- RTE_FLOW_ACTION_TYPE_SET_MAC_SRC
- RTE_FLOW_ACTION_TYPE_SET_MAC_DST
added in order to offload to NIC

The rte_flow_itme_eth must be present in rte_flow pattern

Signed-off-by: Xiaoyu Min <jackmin@mellanox.com>
Acked-by: Yongseok Koh <yskoh@mellanox.com>
Acked-by: Andrew Rybchenko <arybchenko@solarflare.com>

net/mlx5: rewrite TTL by E-Switch

Offload following modify TTL actions to E-Switch via
TC-Flower driver

- RTE_FLOW_ACTION_TYPE_SET_TTL
- RTE_FLOW_ACTION_TYPE_DEC_TTL

The corresponding IP protocol rte_flow_item_ipv[4|6]
must be present in rte_flow pattern otherwith PMD
return error

The example testpmd commands are:

    flow create 0 transfer ingress
         pattern eth / ipv4 / udp dst is 7000 / end
 actions dec_ttl /
 port_id id 1 / end

    flow create 0 transfer ingress
         pattern eth / ipv4 / udp dst is 7001 / end
 actions set_ttl ttl_value 10 /
 port_id id 1 / end

Signed-off-by: Xiaoyu Min <jackmin@mellanox.com>
Acked-by: Yongseok Koh <yskoh@mellanox.com>

app/testpmd: add commands to modify TTL

add commands which supports following TTL actions:
- RTE_FLOW_ACTION_TYPE_DEC_TTL
- RTE_FLOW_ACTION_TYPE_SET_TTL

Signed-off-by: Xiaoyu Min <jackmin@mellanox.com>
Acked-by: Yongseok Koh <yskoh@mellanox.com>

ethdev: add generic TTL rewrite actions

rewrite TTL by decrease or just set it directly
it's not necessary to check if the final result
is zero or not

This is slightly different from the one defined
by openflow and more generic

Signed-off-by: Xiaoyu Min <jackmin@mellanox.com>
Acked-by: Yongseok Koh <yskoh@mellanox.com>
Acked-by: Andrew Rybchenko <arybchenko@solarflare.com>

ethdev: add field for device data per process

Primary and secondary processes share a per-device private data. With
current design it is not possible to have data per-device per-process.
This is required for handling properly the CPP interface inside the NFP
PMD with multiprocess support.

There is also at least another PMD driver, tap, with similar
requirements for per-process device data.

Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
Reviewed-by: Andrew Rybchenko <arybchenko@solarflare.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

doc: add ixgbe known issue with legacy interrrupt

When using uio_pci_generic module or using legacy interrupt mode of
igb_uio or vfio, X550 cannot get interrupts. Because the Interrupt
Status bit is not implemented, then the irq cannot be handled correctly
and cannot report the event fd to DPDK apps.

Add this hw limitation and details into ixgbe known issue.

Signed-off-by: Xiaoyun Li <xiaoyun.li@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>

net/cxgbe: add action to swap MAC addresses

Implement RTE_FLOW_ACTION_TYPE_MAC_SWAP to offload flows with
action to swap the source and destination MAC addresses in the
outermost Ethernet header.

Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>

net/ark: support updated hardware and Tx meta data

Support for version 2 of DDM hardware and user tx meta data.
Verify version during initialization.

Signed-off-by: Ed Czeck <ed.czeck@atomicrules.com>

net/ark: remove useless hardware notifications

Only send notification to mpu when crossing 64 index boundary.

Signed-off-by: Ed Czeck <ed.czeck@atomicrules.com>

net/ark: add recovery for lack of mbufs

Attempt to allocate smaller chunk of mbufs when larger amount is
not available.  Report error when small chunk not available.

Signed-off-by: Ed Czeck <ed.czeck@atomicrules.com>

app/testpmd: fix vdev socket initialization

The cmdline settings of port-numa-config and rxring-numa-config have
been flushed by the following init_config. If we don't configure the
port-numa-config, the virtual device will allocate the device ports to
socket 0. It will cause failure when the socket 0 is unavailable.

eg:
testpmd -l <cores from socket 1> --vdev net_pcap0,iface=lo
--socket-mem=64 -- --numa --port-numa-config="(0,1)"
--ring-numa-config="(0,1,1),(0,2,1)" -i

...
Configuring Port 0 (socket 0)
Failed to setup RX queue:No mempool allocation on the socket 0
EAL: Error - exiting with code: 1
  Cause: Start ports failed

Fix by allocate the devices port to the first available socket or the
socket configured in port-numa-config.

Fixes: 487f9a592a27 ("app/testpmd: fix NUMA structures initialization")
Fixes: 20a0286fd2c0 ("app/testpmd: check socket id validity")
Cc: stable@dpdk.org
Signed-off-by: Phil Yang <phil.yang@arm.com>
Reviewed-by: Gavin Hu <gavin.hu@arm.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/qede: remove invalid config option checks

Now that logging of initialization is controlled by dynamic debug
level, remove the #ifdef for RTE_LIBRTE_QEDE_DEBUG_INIT since that
config option does not exist in common_base and is not documented.

Fixes: 69b6573980c1 ("net/qede: implement dynamic logging")
Cc: stable@dpdk.org
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

doc: remove invalid config options from liquidio guide

The config options for init and debug driver were replaced by
dynamic log levels.

Fixes: 70e28c612d65 ("net/liquidio: implement dynamic logging")
Cc: stable@dpdk.org
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

doc: remove invalid config option from ena guide

The ena driver has used dynamic log type to enable init debugging since
18.05 but the documentation still has leftover mention of the config
option.

Fixes: 8bc0acae970d ("net/ena: implement dynamic logging")
Cc: stable@dpdk.org
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/avf: fix build with debug enabled

DEBUG_DUMP_DESC flag is enabled/disabled from Makefile and enabling it
causing the build issue.

Add the missed 'volatile' keyword to avoid the warning for type
mismatch, which will be treated as compiler error if WERROR_FLAGS
is enabled.

Fixes: bfd38e4d708b ("net/avf: fix missing compiler error flags")
CC: stable@dpdk.org
Signed-off-by: Haiyue Wang <haiyue.wang@intel.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>

net/mlx5: fix build with zero-size array

If the build environment doesn't have 'linux/tc_act/tc_pedit.h' header,
compiler will use needed structs defined in mlx5_flow_tcf.c.

However, there is a zero-size array defined in one struct and
ISO C forbids this when -Wpedantic is set by debug mode.

Simply put __extension__ keyword before the struct in question.

Fixes: 2ed2fe5f0a9c ("net/mlx5: rewrite IP address UDP/TCP port by E-Switch")

Signed-off-by: Xiaoyu Min <jackmin@mellanox.com>

net/mlx5: support multiple groups and jump action

rte_flow has 'group' attribute and 'jump' action in order to support
multiple groups. This feature is known as multi-table support ('chain' in
linux TC flower) in general because a group means a table of flows. Example
commands are:

flow create 0 transfer priority 1 ingress
     pattern eth / vlan vid is 100 / end
     actions jump group 1 / end

flow create 0 transfer priority 1 ingress
     pattern eth / vlan vid is 200 / end
     actions jump group 2 / end

flow create 0 transfer group 1 priority 2 ingress
     pattern eth / vlan vid is 100 /
           ipv4 dst spec 192.168.40.0 dst prefix 24 / end
     actions drop / end

flow create 0 transfer group 1 priority 2 ingress
     pattern end
     actions of_pop_vlan / port_id id 1 / end

flow create 0 transfer group 2 priority 2 ingress
     pattern eth / vlan vid is 200 /
           ipv4 dst spec 192.168.40.0 dst prefix 24 / end
     actions of_pop_vlan / port_id id 2 / end

flow create 0 transfer group 2 priority 2 ingress
     pattern end
     actions port_id id 2 / end

With theses flows, if a packet having vlan 200 and src_ip as 192.168.40.1,
this packet will firstly hit the 1st flow. Then it will hit the 5th flow
because of the 'jump' action. As a result, the packet will be forwarded to
port 2 (VF representor) with vlan tag being stripped off. If the packet had
vlan 100 instead, it would be dropped by the 3rd flow.

Signed-off-by: Yongseok Koh <yskoh@mellanox.com>

net/atlantic: add release notes

Release notes for Aquantia atlantic driver.

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>

net/atlantic: read MAC registers for debug purposes

This patch add support for dumping MAC registers.

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>

net/atlantic: implement EEPROM get/set

Add support for EEPROM reading/writing.

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>

net/atlantic: implement VLAN filters and offloads

VLAN filters and VLAN offloads implementation.

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>

net/atlantic: implement MTU configuration

Add support for updating MTU value.

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>

net/atlantic: implement MAC address manipulations

Add support for adding/removing MAC addresses.

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>

net/atlantic: add flow control configuration

Add support for flow control feature.

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>

net/atlantic: implement RSS and RETA manipulation API

Add support for Receive Side Scaling feature.
RSS hash and reta table configuration.

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>

net/atlantic: implement promisc and allmulti modes

Add support for promiscuous/allmulticast modes configuration.

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>

net/atlantic: add Rx/Tx descriptors information

Add support for Rx/Tx descriptors status information.

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>

net/atlantic: implement device statistics

Add support for device stats, xstats and queue stats

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>

net/atlantic: add link status and interrupt management

Implement link interrupt, link info, link polling.

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>

net/atlantic: implement Tx path

Add implementation for TX datapath.

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>

net/atlantic: implement Rx path

Add implementation for RX datapath.

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>

net/atlantic: implement core logic for Tx/Rx

Add RX/TX function prototypes for further datapath development.

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>

net/atlantic: configure device start/stop

Start, stop and reset are all done via hw_atl layer.

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>

net/atlantic: add b0 hardware layer

This is hw_atl logic layer derived from linux atlantic
driver. It contains RX/TX hardware initialization
sequences, various hw configuration.

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>

net/atlantic: implement firmware operations

AQC NICs comes in fields with two major
FW generations: 1x and 3x.

This is part of linux atlantic driver shared code,
responsible for internal NIC firmware interactions,
including link management ops, FW initialization,
various lifecycle features.

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>

net/atlantic: add hardware access layer

This patch introduces hw_atl layer which is maintained
in sync with linux kernel driver.
It will generally be in sync with linux upstream.

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>

net/atlantic: add hardware registers access routines

Add implementation for hardware registers access routines.

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>

net/atlantic: add logging structure

Implement logging macroses for debug purposes.

Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>