The rte_compat.h header file must be included to get the definition of
__rte_experimental.
Fixes: f030bff72f ("bitrate: add free function")
Cc: stable@dpdk.org
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: David Marchand <david.marchand@redhat.com>
The rte_mbuf_dyn.h header file uses a number of types and macros without
including the required header files to get the definitions of those
macros/types. Similarly, the rte_mbuf_core.h file was missing an
include for rte_byteorder.h header.
Fixes: 4958ca3a44 ("mbuf: support dynamic fields and flags")
Fixes: 3eb860b08e ("mbuf: move definitions into a separate file")
Cc: stable@dpdk.org
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Reviewed-by: David Marchand <david.marchand@redhat.com>
The define for RTE_ETH_FLOW_MAX is defined in rte_ethdev.h, so that
header should be included in rte_eth_ctrl.h to allow it to be compiled
independently.
Fixes: 7fa96d696f ("ethdev: unification of flow types")
Cc: stable@dpdk.org
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Reviewed-by: David Marchand <david.marchand@redhat.com>
The telemetry header file uses the rte_cpuset_t type, but does not
include any header providing that type. Include rte_os.h to provide the
necessary type.
Fixes: febbebf7f2 ("telemetry: keep threads separate from data plane")
Cc: stable@dpdk.org
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: David Marchand <david.marchand@redhat.com>
The rte_reciprocal header file used standard __rte_always_inline from
rte_common.h but does not include that header file, leading to compiler
errors when the reciprocal header is included alone.
Fixes: 6d45659eac ("eal: add u64-bit variant for reciprocal divide")
Cc: stable@dpdk.org
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: David Marchand <david.marchand@redhat.com>
rte_thread.h was missing the compat header to get the __rte_experimental
macro definition.
Fixes: b1fd151267 ("eal: add generic thread-local-storage functions")
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: David Marchand <david.marchand@redhat.com>
Rather than having the DPDK configuration error out when linking apps
and examples when "both" is select for "default_library" option, we can
detect that setting earlier in the build config and provide a suitable
error message to the user.
Cc: stable@dpdk.org
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: Andrew Boyer <aboyer@pensando.io>
Reviewed-by: Honnappa Nagarahalli <honnappa.nagarahalli@arm.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
As announced in the deprecation note, remove all compatibility build
defines from previous make/meson versions and use only the standardized
ones - RTE_LIB_<name> for libraries, and RTE_<CLASS>_<NAME> for drivers.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Add empty item into the pattern mapping list when checking the ptype is
supported or not, since we do support the NUll pattern for rte flow in
ice pmd.
Fixes: 68b6240ee9 ("net/ice: refactor packet type parsing")
Signed-off-by: Jeff Guo <jia.guo@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
Currently, IXGBE doesn't support RSS when double VLAN
is enabled due to FW limitation.
Signed-off-by: Beilei Xing <beilei.xing@intel.com>
Acked-by: Haiyue Wang <haiyue.wang@intel.com>
For some types of NIC, jumbo frame is not supported in IOV mode,
so if a VF requests to configure the frame size to not bigger
than IXGBE_ETH_MAX_LEN, the kernel driver returns 0, but the DPDK
ixgbe PMD returns -1, this will cause the VF to fail to start
when the PF driven by DPDK ixgbe PMD.
This patch keeps ixgbe PMD's handling mode consistent with kernel
driver in above situation.
In addition, the value set by the command IXGBE_VF_SET_LPE
represents the max frame size, not the mtu.
Fixes: 1b9ea09c06 ("ixgbe: support X550")
Fixes: 95a27b3ba5 ("net/ixgbe: enable jumbo frame for VF")
Cc: stable@dpdk.org
Signed-off-by: Alvin Zhang <alvinx.zhang@intel.com>
Acked-by: Jeff Guo <jia.guo@intel.com>
Since DCF always configure the outer VLAN offloads for the target AVF,
so rename the related variables to align with this design.
Also, the DCF needs to trace the AVF reset status to re-apply the VLAN
offload setting, refactor the reset event handling to support this.
Change the VF representor API 'ethdev' parameter to 'vf_rep_eth_dev' to
avoid introducing confusion with VF representor eth_dev ops name.
Signed-off-by: Haiyue Wang <haiyue.wang@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
Allow to ignore error returned from virtual channel command
VIRTCHNL_OP_SET_RSS_HENA with hena = 0 due to some legacy
kernel driver does not support this. Add warning message when
this happen.
Signed-off-by: Xuan Ding <xuan.ding@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
PMD validates the rss action in the sample sub-actions list,
then translates into rdma-core action and it will be used for sample
path destination.
If the RSS action is in both sample sub-actions list and original flow,
the rss level and rss type in the sample sub-actions list should be
consistent with the original flow list, since the expanding items
for RSS should be the same for both actions.
Signed-off-by: Jiawei Wang <jiaweiw@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
RSS action is valid only in NIC-RX domain, this fix bypass
the function that getting RSS action from the flow action list
under no NIC-RX domain.
Fixes: e745f90007 ("net/mlx5: optimize flow RSS struct")
Cc: stable@dpdk.org
Signed-off-by: Jiawei Wang <jiaweiw@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
Support rss action in the sample sub-actions list.
The examples for the sample flow use case and result as below:
set sample_actions 0 mark id 0x12 / rss queues 0 1 2 3 end / end
flow create 0 ingress group 1 pattern eth / end actions
sample ratio 1 index 0 / jump group 2 / end
This flow will result in all the matched ingress packets will be
jumped to next table, and the each packet will be marked with 0x12
and duplicated to rss queues of the control application.
Signed-off-by: Jiawei Wang <jiaweiw@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
An application may submit all-zeros masks for a given field
to be used in two given specifications and, in the meantime,
use different unmasked values. Later on, when compared, the
two specifications will prove unequal, and, if the rules in
question are outer, the client driver will have to allocate
a separate rule for the second specification. Provided that
all other match criteria are the same, the HW will deem the
two outer rules being duplicates, which is in fact the case.
Apply masks to values in efx_mae_match_spec_field_set() API
to fix the issue and avoid duplicate outer rule allocations.
Fixes: 370ed675a9 ("common/sfc_efx/base: support setting PPORT in match spec")
Cc: stable@dpdk.org
Reported-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Signed-off-by: Ivan Malov <ivan.malov@oktetlabs.ru>
Reviewed-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Reviewed-by: Andy Moreton <amoreton@xilinx.com>
Add support for the RTE_FLOW_ACTION_MODIFY_FIELD to the testpmd.
Implement CLI to create the modify_field action and supply all the
needed parameters to modify an arbitrary packet field (as well as
mark, tag or metadata) with data from another field or immediate
value.
Example of the flow is the following:
flow create 0 egress group 1 pattern eth / ipv4 / udp / end
actions modify_field op set dst_type tag dst_level 2 dst_offset 8
src_type gtp_teid src_level 0 src_offset 0 width 16 / end
This flow copies 16 bits from the second Tag in the Tags array
into the outermost GTP TEID packet header field. 8 bits of the
Tag are skipped as indicated by the dst_offset action parameter.
op, dst_type, src_type and width are the mandatory parameters to
specify. Levels and offset are 0 by default if they are not
overridden by a user. The operation can be set, add or sub.
Signed-off-by: Alexander Kozyrev <akozyrev@nvidia.com>
Acked-by: Ori Kam <orika@nvidia.com>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Implement the generic modify flow API to allow manipulations on
an arbitrary header field (as well as mark, metadata or tag) using
data from another field or a user-specified value.
This generic modify mechanism removes the necessity to implement
a separate RTE Flow action every time we need to modify a new packet
field in the future.
Supported operation are:
- set: copy data from source to destination.
- add: integer addition, stores the result in destination.
- sub: integer subtraction, stores the result in destination.
The field ID is used to specify the desired source/destination packet
field in order to simplify the API for various encapsulation models.
Specifying the packet field ID with the needed encapsulation level
is able to quickly get a packet field for any inner packet header.
Alternatively, the special ID (ITEM_START) can be used to point to
the very beginning of a packet. This ID in conjunction with the
offset parameter provides great flexibility to copy/modify any part of
a packet as needed.
The number of bits to use from a source as well as the offset can be
be specified to allow a partial copy or dividing a big packet field
into multiple small fields (e.g. copying 128 bits of IPv6 to 4 tags).
An immediate value (or a pointer to it) can be specified instead of the
level and the offset for the special FIELD_VALUE ID (or FIELD_POINTER).
Can be used as a source only.
Signed-off-by: Alexander Kozyrev <akozyrev@nvidia.com>
Acked-by: Ori Kam <orika@nvidia.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
The default RSS key does not need to be configured repeatedly
when call hns3_dev_configure function with the NULL RSS key
because the default RSS key has been configured when the PMD
driver run hns3_do_start function with starting device.
Besides, it will not overwrite the initialized key if
rte_eth_dev_configure API will be called directly and RSS key is NULL
after init PMD driver.
Therefore, the assignment for RSS key in hns3_dev_configure
function is unnecessary.
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Currently, when dump the queue interrupt registers, the number of
registers that should be dumped is calculated from num_msi. But the
value of num_msi includes the number of misc interrupts. So, for some
hardware version, like kupeng930, it will lead to an illegal access.
This patch replace num_msi with intr_tqps_num which indicate the
number of interrupts used by the tqps.
Fixes: 936eda25e8 ("net/hns3: support dump register")
Cc: stable@dpdk.org
Signed-off-by: Chengchang Tang <tangchengchang@huawei.com>
Signed-off-by: Lijun Ou <oulijun@huawei.com>
The data pointer has not moved after BAR register dumped. This causes
the later register to overwrite the previous data.
This patch fix the overwriting by move the pointer after every dump
function. And the missing separator between 32-bit register and the
64-bit register is also added to avoid a parsing error.
Fixes: 936eda25e8 ("net/hns3: support dump register")
Cc: stable@dpdk.org
Signed-off-by: Chengchang Tang <tangchengchang@huawei.com>
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Currently, the reg length return by HNS3 is the total length of all the
registers. But for upper layer user, the total register length is the
length multiplied by width. This can lead to a waste of memory and print
some invalid information.
This patch corrects the length and width of the register.
Fixes: 936eda25e8 ("net/hns3: support dump register")
Cc: stable@dpdk.org
Signed-off-by: Chengchang Tang <tangchengchang@huawei.com>
Signed-off-by: Lijun Ou <oulijun@huawei.com>
The original command opcode '0x700A' may cause firmware error,
so '0x700A' is deserted, now use '0x700B' to replace it.
Fixes: 223d9eceae ("net/hns3: clear residual hardware configurations on init")
Cc: stable@dpdk.org
Signed-off-by: Hongbo Zheng <zhenghongbo3@huawei.com>
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Current hns3vf queried link status in dev init stage, but the link
status should be maintained in dev start stage, this patch fix this.
Also, in the dev start stage, we use quick query instead of delayed
query to make sure update the link status soon.
Fixes: a5475d61fa ("net/hns3: support VF")
Fixes: 958edf6627 ("net/hns3: fix VF link status")
Cc: stable@dpdk.org
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Lijun Ou <oulijun@huawei.com>
This patch support RTE TM ops function for PF, which could
used to:
1. config port's peak rate.
2. config TC's peak rate.
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Use C11 atomic builtins with explicit ordering instead of
rte_atomic ops with the resetting member of hns3_reset_data
structure.
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Don't assume autoneg in link_update().
Always call ionic_dev_cmd_port_autoneg() in start().
This allows the client to specify the link settings.
Fixes: 598f672639 ("net/ionic: add basic port operations")
Cc: stable@dpdk.org
Signed-off-by: Andrew Boyer <aboyer@pensando.io>
The address is not needed in the firmware.
This reduces the startup time a bit.
Signed-off-by: Andrew Boyer <aboyer@pensando.io>
Signed-off-by: Neel Patel <neel@pensando.io>
Adding F_ENA to the q_init command has the same effect as q_enable.
This reduces the startup time a bit.
Signed-off-by: Andrew Boyer <aboyer@pensando.io>
Signed-off-by: Neel Patel <neel@pensando.io>
Configuration flags come to the driver in dev_configure(). From there,
the driver calls ionic_lif_configure() to update the lif->feature
bitfield, and then programs the port.
Features like VLAN_RX_FILTER and RX_HASH cannot be disabled in the
device, so do nothing in response to requests to disable them. (The
device config would ideally show them enabled by default, but some
DTS tests fail if RSS_HASH is set but RSS is not.)
Move features from the per-queue to per-port lists. IONIC does not
really support per-queue configuration: the stack disallows disabling
a queue feature if it is enabled on the port, while the device
disallows enabling a queue feature if it is disabled on the port.
Thus all configuration is per-port.
Move the guts of ionic_vlan_offload_set() into a new function,
ionic_lif_configure_vlan_offload(), so it can be called by
ionic_lif_configure().
Move the check for DEV_RX_OFFLOAD_SCATTER from rx_queue_setup() up
into ionic_lif_configure().
Warn if rx_drop_en is not set.
Signed-off-by: Andrew Boyer <aboyer@pensando.io>
Strict-aliasing rules are violated by cast to uint16_t* in flowgen.c and
the calculated IP checksum is wrong. Use attribute __may_alias__ to fix
the problem.
Fixes: e9e23a617e ("app/testpmd: add flowgen forwarding engine")
Cc: stable@dpdk.org
Signed-off-by: George Prekas <prekageo@amazon.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
Latest VIC adapters support 64B CQ (completion queue) entries as well
as 16B entries available on all VIC models. 64B entries can greatly
reduce cache contention (CPU stall cycles) between DMA writes (Rx
packet descriptors) and polling CPU. The effect is very noticeable on
Intel platforms with DDIO. As most UCS servers are based on Intel
platforms, enable and use 64B CQ entries by default, if
available. Also, add devarg 'cq64' so the user can explicitly disable
64B CQ.
Signed-off-by: Hyong Youb Kim <hyonkim@cisco.com>
Reviewed-by: John Daley <johndale@cisco.com>
When querying the link status via telemetry interface, we don't want the
client to have to wait for multiple seconds for a reply. Therefore use
"rte_eth_link_get_nowait()" rather than "rte_eth_link_get()" in the
telemetry callback.
Fixes: c190daedb9 ("ethdev: add telemetry callbacks")
Cc: stable@dpdk.org
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Ciara Power <ciara.power@intel.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
This config item is not exposed anymore now that we removed make
support.
Signed-off-by: David Marchand <david.marchand@redhat.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
This config item is not exposed anymore now that we removed make
support.
Note: all architectures provide vectorised functions.
Signed-off-by: David Marchand <david.marchand@redhat.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
Following make support removal, the vectorised code is not built
anymore, fix the build flag check.
Fixes: 3cc6ecfdfe ("build: remove makefiles")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
The GENEVE TLV option matching flows must be created
using a translation function.
This function checks whether we already created a Devx
object for the matching and either creates the objects
or updates the reference counter.
Signed-off-by: Shiri Kuzin <shirik@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
This patch adds validation routine for the GENEVE
header TLV option.
The GENEVE TLV option match must include all fields
with full masks due to NIC does not support masking
on option class, type and length.
The option data length must be non zero and provided
data pattern should be zero neither due to hardware
limitations.
Signed-off-by: Shiri Kuzin <shirik@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
Currently firmware supports the only TLV object per device
to match on the GENEVE header option.
This patch adds the simple TLV object management to the mlx5 PMD.
Signed-off-by: Shiri Kuzin <shirik@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
TLV object is a special firmware maintained entity used
to support match on GENEVE header extension option.
The TLV object is created with DevX API and accepts
the option class, type and lehgth fields.
The class type and length fields are set using MLX5_SET
and the Devx object is created using mlx5 glue function.
Signed-off-by: Shiri Kuzin <shirik@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
This is preparation step to support match on GENEVE TLV option.
In this Patch we add the HCA attributes that will allow
supporting GENEVE TLV option matching.
Signed-off-by: Shiri Kuzin <shirik@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
The patch adds the GENEVE rte flow option length support to
command line interpreter. The flow command with GENEVE
option items looks like:
flow create 0 ingress pattern eth / ipv4 / udp / geneve vni
is 100 optlen is 2 / end actions drop / end
The option length should be specified in 32-bit words, this
value specifies the all options length in the GENEVE header.
Signed-off-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
Acked-by: Ori Kam <orika@nvidia.com>
The patch adds the GENEVE option rte flow item support to
command line interpreter. The flow command with GENEVE
option items looks like:
flow create 0 ingress pattern eth / ipv4 / udp / geneve vni is 100 /
geneve-opt class is 99 length is 1 type is 0 data is 0x669988 /
end actions drop / end
The option length should be specified in 32-bit words, this
value specifies the length of the data pattern/mask arrays (should be
multiplied by sizeof(uint32_t) to be expressed in bytes. If match
on the length itself is not needed the mask should be set to zero, in
this case length is used to specify the pattern/mask array lengths only.
Signed-off-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
Acked-by: Ori Kam <orika@nvidia.com>
