The memory barrier was intended for descriptor data integrity (see
comments in [1]). As later NEON loads were implemented and a whole
entry is loaded in one-run and atomic, that makes the ordering of
partial loading unnecessary. Remove it accordingly.
Corrected couple of code comments.
In terms of performance, observed slightly higher average throughput
in tests with 82599ES NIC.
[1] http://patches.dpdk.org/patch/18153/
Fixes: 989a840505 ("net/ixgbe: fix received packets number for ARM NEON")
Cc: stable@dpdk.org
Signed-off-by: Ruifeng Wang <ruifeng.wang@arm.com>
Reviewed-by: Gavin Hu <gavin.hu@arm.com>
As packet length extraction code was simplified,the ordering
was not necessary any more.[1]
2% performance gain was measured on Marvell ThunderX2.
4.3% performance gain was measured on Ampere eMAG80
[1] http://mails.dpdk.org/archives/dev/2016-April/037529.html
Fixes: ae0eb310f2 ("net/i40e: implement vector PMD for ARM")
Cc: stable@dpdk.org
Signed-off-by: Gavin Hu <gavin.hu@arm.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
Reviewed-by: Steve Capper <steve.capper@arm.com>
For x86, the descriptors needs to be loaded in order, so in between two
descriptors loading, there is a compiler barrier in place.[1]
For aarch64, a patch [2] is in place to survive with discontinuous DD
bits, the barriers can be removed to take full advantage of out-of-order
execution.
50% performance gain in the RFC2544 NDR test was measured on ThunderX2.
12.50% performance gain in the RFC2544 NDR test was measured on Ampere
eMAG80 platform.
[1] http://inbox.dpdk.org/users/039ED4275CED7440929022BC67E7061153D71548@
SHSMSX105.ccr.corp.intel.com/
[2] https://mails.dpdk.org/archives/stable/2017-October/003324.html
Fixes: ae0eb310f2 ("net/i40e: implement vector PMD for ARM")
Cc: stable@dpdk.org
Signed-off-by: Gavin Hu <gavin.hu@arm.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
Reviewed-by: Steve Capper <steve.capper@arm.com>
The AltiVec header file breaks boolean type:
error: incompatible types when initializing type
'__vector _bool int' {aka '_vector(4) __bool int'} using type 'int'
If __APPLE_ALTIVEC__ is defined, then bool type is redefined
and conflicts with stdbool.h.
There is no good solution to fix it for the whole project without
breaking something else, so a workaround is inserted in mlx5 PMD.
This workaround is not compatible with C++ but there is no C++ in DPDK.
Related to: 725f5dd0bf ("net/mlx5: fix build on PPC64")
Cc: stable@dpdk.org
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
Tested-by: David Christensen <drc@linux.vnet.ibm.com>
Acked-by: Matan Azrad <matan@mellanox.com>
atl_dev_info_get() is declared twice in atl_ethdev.c.
Delete one of these declarations.
Fixes: bb42aa9ffe ("net/atlantic: configure device start/stop")
Cc: stable@dpdk.org
Signed-off-by: Ivan Ilchenko <ivan.ilchenko@oktetlabs.ru>
Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>
Acked-by: Igor Russkikh <igor.russkikh@aquantia.com>
Add multiple processes support for ice, secondary processes will share
memory and configuration with primary process, do not need further
initialization for secondary processes.
Signed-off-by: Xiao Zhang <xiao.zhang@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
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.
Signed-off-by: Rastislav Cernay <cernay@netcope.com>
Acked-by: Jan Remes <remes@netcope.com>
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.
Signed-off-by: Rastislav Cernay <cernay@netcope.com>
Reviewed-by: Xiaolong Ye <xiaolong.ye@intel.com>
Acked-by: Jan Remes <remes@netcope.com>
The testpmd application provides two sets of commands for RX offload
flags configuration. The purpose of this patch is to eliminate this
duplication by removing the old set of commands:
“port config all crc-strip|scatter|rx-cksum|rx-timestamp|hw-vlan|
hw-vlan-filter|hw-vlan-strip|hw-vlan-extend on|off”
The other commands set that can be used instead in order to enable
or disable the same RX offloading flags on all RX queues of a port is:
"port config <port_id> rx_offload crc_strip|scatter|ipv4_cksum|
udp_cksum|tcp_cksum|timestamp|vlan_strip|vlan_filter|vlan_extend on|off"
This patch also fixes the "drop-en" command, which enables packets
dropping on all RX queues of all ports when no receive buffers available
“port config all drop-en on|off”
Fixes: 384161e006 ("app/testpmd: adjust on the fly VLAN configuration")
Cc: stable@dpdk.org
Signed-off-by: Flavia Musatescu <flavia.musatescu@intel.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
af_packet driver is leaving stale socket after device is removed.
Ring buffers are memory mapped when device is added using rte_dev_probe.
There is no corresponding munmap call when device is removed/closed.
This commit fixes the issue by calling munmap
from rte_pmd_af_packet_remove().
Bugzilla ID: 339
Fixes: e6ee4db01b ("af_packet: make the device detachable")
Cc: stable@dpdk.org
Signed-off-by: Abhishek Sachan <abhishek.sachan@altran.com>
Reviewed-by: John W. Linville <linville@tuxdriver.com>
Security Parameters Index (SPI) should be set with network endian
values.
While 0xffffffff == htonl(0xffffffff), this missing annotation is
caught by sparse when compiling ovs (dpdk-latest branch).
Fixes: d4b684f719 ("net: add ESP header to generic flow steering")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
Reviewed-by: Andrew Rybchenko <arybchenko@solarflare.com>
There is no RTE_FDIR_DISABLE. The right name is RTE_FDIR_MODE_NONE.
Fixes: af75078fec ("first public release")
Cc: stable@dpdk.org
Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>
Timestamp is always set in PCAP header, whether it reads a file or
listen on an interface. This information can be important for some
applications and it cannot be obtained otherwise (especially when
reading a PCAP file, where the timestamp is not the current time).
Timestamp here is the number of microseconds since epoch.
Signed-off-by: Sylvain Rodon <srn@nexatech.fr>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
The generic RTE_LOGTYPE_PMD is a historical relic and should
be deprecated. Every driver must register its own logtype.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
The generic RTE_LOGTYPE_PMD is a historical relic and should
be deprecated. Every driver must register its own logtype.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
Replace the boilerplate BSD license text with the equivalent
SPDX license id.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Tetsuya Mukawa <mtetsuyah@gmail.com>
Acked-by: Takanari Hayama <taki@igel.co.jp>
Fix some mistakes in Tx bursts in regard to TSO flag check.
Fixes: 18a1c20044 ("net/mlx5: implement Tx burst template")
Cc: stable@dpdk.org
Signed-off-by: Yongseok Koh <yskoh@mellanox.com>
For the rte_flow_action_raw_encap, the header definition for
encapsulation must be available, otherwise it will lead to crash on some
OFED versions and logically it should be rejected.
Fixes: 8ba9eee4ce ("net/mlx5: add raw data encap/decap to Direct Verbs")
Cc: stable@dpdk.org
Signed-off-by: Xiaoyu Min <jackmin@mellanox.com>
Acked-by: Dekel Peled <dekelp@mellanox.com>
The ops pointers in fm10k_stats_get() are set up from primary
process, when secondary process calls these ops pointers,
a segment fault will happen.
Fixes: 7223d200c2 ("fm10k: add base driver")
Cc: stable@dpdk.org
Signed-off-by: Lu Qiuwen <luqiuwen@iie.ac.cn>
Acked-by: Xiao Wang <xiao.w.wang@intel.com>
Reviewed-by: Xiaolong Ye <xiaolong.ye@intel.com>
I40evf queue can not work properly with kernel pf driver for X722 vf.
Eg. when configure 8 queues pair, only 4 queues can receive packets,
and half packets will be lost if using 2 queues pair.
This issue is caused by misconfiguration of look up table, the original
code of LUT configuration did not work for X722 vf, use aq command to
setup the LUT to make it work properly.
Fixes: cea7a51c17 ("i40evf: support RSS")
Cc: stable@dpdk.org
Acked-by: Beilei Xing <beilei.xing@intel.com>
Signed-off-by: Xiao Zhang <xiao.zhang@intel.com>
Rather than the global promiscuous mode on all slaves, let's try to use
allmulti.
To do this, we apply the same mechanism than for promiscuous and drop in
rx_burst.
When adding a slave, we first try to use allmulti, else we try
promiscuous.
Because of this, we must be able to handle allmulti on the bonding
interface, so the associated dev_ops is added with checks on which mode
has been applied on each slave.
Rather than add a new flag in the bond private structure, we can remove
promiscuous_en since ethdev already maintains this information.
When starting the bond device, there is no promisc/allmulti re-apply
as, again, ethdev does this itself.
On reception, we must inspect if the packets are multicast or unicast
and take a decision to drop based on which mode has been enabled on the
bonding interface.
Note: in the future, if we define an API so that we can add/remove mc
addresses to a port (rather than the current global set_mc_addr_list
API), we can refine this to only register the LACP multicast mac
address.
Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Chas Williams <chas3@att.com>
By default, the 802.3ad code enables promisc mode on all slaves.
To avoid all packets going to the application (unless the application
asked for promiscuous mode), all frames are supposed to be filtered in
the rx burst handler.
However the incriminated commit broke this because non pure ethernet
frames (basically any unicast Ether()/IP() packet) are not filtered
anymore.
Fixes: 71b7b37ec9 ("net/bonding: use ptype flags for LACP Rx filtering")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Chas Williams <chas3@att.com>
Fast queue Rx burst function is missing checks on promisc and the
slave collecting state.
Define an inline wrapper to have a common base.
Fixes: 112891cd27 ("net/bonding: add dedicated HW queues for LACP control")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Chas Williams <chas3@att.com>
We'd better consolidate the fast queue and the normal tx burst functions
under a common inline wrapper for maintenance.
But looking closer at the bufs_slave_port_idxs[] mapping array in those
tx burst functions, its size is invalid since up to nb_bufs are handled
here.
A previous patch [1] fixed this issue for balance tx burst function
without mentioning it.
802.3ad and balance modes are functionally equivalent on transmit.
The only difference is on the slave id distribution.
Add an additional inline wrapper to consolidate even more and fix this
issue.
[1]: https://git.dpdk.org/dpdk/commit/?id=c5224f623431
Fixes: 09150784a7 ("net/bonding: burst mode hash calculation")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Chas Williams <chas3@att.com>
Start a new release cycle with empty release notes.
Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
This patch for DPDK Contributor's Guidelines indicates the repos
against which a new PMD should be prepared; for example, for new
network ethernet PMDs it should be dpdk-next-net, and for new crypto
PMDs it should be dpdk-next-crypto. For other new PMDs, the
contributor should refer to the MAINTAINERS file. Though this may seem
obvious, it is not mentioned in DPDK documentation.
Signed-off-by: Rami Rosen <ramirose@gmail.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
Last contribution, last commit closing the release.
Thank you Rami, you will be missed.
In APP_MODE_SRTCM_COLOR_AWARE mode, sample app compilation fails
due to wrong meter color type conversion.
error log-
/qos_meter/main.c:error: conversion to incomplete type
(enum rte_meter_color) input_color);
Fixes: c1656328db ("meter: replace color definitions")
Cc: stable@dpdk.org
Reported-by: Yuan Peng <yuan.peng@intel.com>
Signed-off-by: Jasvinder Singh <jasvinder.singh@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
This reverts commit aa2c00702b, which
introduced the possibility of an invalid address exception when running
an application with a stopped receive queue. The issues with rxq stop/start
will be revisited in the 19.11 release timeframe.
Fixes: aa2c00702b ("net/bnxt: fix traffic stall on Rx queue stop/start")
Cc: stable@dpdk.org
Signed-off-by: Lance Richardson <lance.richardson@broadcom.com>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
It may be needed by the user to limit the LRO session packet size.
In order to allow the above limitation, a new Rx configuration may be
added for the maximum LRO session size.
A new capability may be added too to expose the maximum LRO session size
supported by the port.
Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Andrew Rybchenko <arybchenko@solarflare.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
Add new offload flags ``DEV_RX_OFFLOAD_RSS_HASH`` and
``DEV_RX_OFFLOAD_FLOW_MARK``.
Add new function ``rte_eth_dev_set_supported_ptypes`` to allow application
to set specific ptypes to be updated in ``rte_mbuf::packet_type``
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Acked-by: Andrew Rybchenko <arybchenko@solarflare.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
The function rte_eth_dev_count() was marked as deprecated in DPDK 18.05
in commit d9a42a69fe ("ethdev: deprecate port count function").
It is planned to be removed after the next LTS release.
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: David Marchand <david.marchand@redhat.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Andrew Rybchenko <arybchenko@solarflare.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
void return value is bad for get API (like rte_eth_dev_info-get())
since caller does not know if the function does its job or not and
output value is filled in.
void return value is bad for state changing API (like
rte_eth_promiscuous_enable()) since caller should use get API
to understand if state is really changed.
Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Tell users about upcoming changes to rte_ether_addr and
rte_ether_header.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Andrew Rybchenko <arybchenko@solarflare.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
New accessors have been introduced to provide the hidden information.
This symbol can now be kept internal.
Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Acked-by: Andrew Rybchenko <arybchenko@solarflare.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
This symbol has been deprecated for quite some time.
Let's drop it as soon as possible.
Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Anatoly Burakov <anatoly.burakov@intel.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Acked-by: Andrew Rybchenko <arybchenko@solarflare.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
This symbol has been deprecated for quite some time.
Let's drop it as soon as possible.
Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Aaron Conole <aconole@redhat.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Andrew Rybchenko <arybchenko@solarflare.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
This commit removes the unnecesarry ast.literal_eval() function call
from the input handling, which now relies just on raw_input() to get
its input.
Fixes: 53f293c9a7 ("usertools: replace unsafe input function")
Cc: stable@dpdk.org
Signed-off-by: Andrius Sirvys <andrius.sirvys@intel.com>
Acked-by: Harry van Haaren <harry.van.haaren@intel.com>
ARM is supporting maximum 4 hugepage sizes (64K, 2M, 32M
and 1G) when granule is 4KB since very long and DPDK
support maximum 3 hugepage sizes.
With all 4 hugepage sizes enabled, applications and some
stacks like VPP which are working over DPDK and using
"in-memory" eal option, or using separate mount points
on ARM based platform, fails at huge page initialization,
reporting error messages from eal:
EAL: FATAL: Cannot get hugepage information.
EAL: Cannot get hugepage information.
EAL: Error - exiting with code: 1
This issue is originated from Linux 5.0
(a21b0b78eaf7 "arm64: hugetlb: Register hugepages during arch init")
where kernel is by default creating directories for each supported
hugepage size in /sys/kernel/mm/hugepages/
On earlier Stable Kernel LTR's, the directories visible in
/sys/kernel/mm/hugepages/ were dependent upon what hugepage
sizes are configured at boot time.
This change increases the maximum supported hugepage sizes
to 4 for ARM based platforms.
Cc: stable@dpdk.org
Signed-off-by: Gagandeep Singh <g.singh@nxp.com>
Signed-off-by: Nipun Gupta <nipun.gupta@nxp.com>
The "and" condition offset == 0 && offset == NVM_INVALID_PTR
can never be true.
Fixes: cf3af5aa56 ("net/ixgbe/base: add functions to get version info")
Cc: stable@dpdk.org
Signed-off-by: Congwen Zhang <zhang.congwen@zte.com.cn>
Acked-by: Xiaolong Ye <xiaolong.ye@intel.com>
The copying of sent mbufs pointers might be deferred to the end of
tx_burst() routine to be copied in one call of rte_memcpy.
For the multi segment packets this optimization is not applicable,
because number of packets does not match with number of mbufs and
we do not have linear array of pointers in pkts parameter.
The completion request generating routine wrongly took into account
the inconsistent (for multi-segment packets) deferred pointer copying.
Fixes: 5a93e173b8 ("net/mlx5: fix Tx completion request generation")
Signed-off-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Introduces the possible limitations on maximal Tx queue
size in descriptors if Tx inline data are enabled.
Signed-off-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
This patch updates mlx5 documentation in parts:
- txq_inline_min parameter is described in more details,
values are fixed
- maximal amount of segments in multi-segment packets.
Fixes: 38b4b397a5 ("net/mlx5: add Tx configuration and setup")
Signed-off-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Acked-by: Raslan Darawsheh <rasland@mellanox.com>
It's possible that the mcsunlock occurs before the test_mcslock_try has
a chance to execute, which will result in the trylock being successful,
making the test case fail. Fix this by waiting until all lcores have
completed their test before unlocking the master lock.
Fixes: 32dcb9fd2a ("test/mcslock: add MCS queued lock unit test")
Signed-off-by: Aaron Conole <aconole@redhat.com>
Acked-by: Phil Yang <phil.yang@arm.com>
Reviewed-by: David Marchand <david.marchand@redhat.com>
On receiving signal, testpmd showing warning as
"LATENCY_STATS: failed to remove Rx/Tx callback"
because rte_latencystats_uninit is called without
checking if latencystats is enabled or not.
After this fix, rte_latencystats_uninit will be
called only if latencystats is enabled.
Fixes: 62d3216d61 ("app/testpmd: add latency statistics calculation")
Cc: stable@dpdk.org
Signed-off-by: Amit Gupta <agupta3@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Tested-by: Jerin Jacob <jerinj@marvell.com>
gcc-9 is stricter on NULL arguments for printf.
Fix the following build error by avoiding NULL argument to printf.
In file included from drivers/net/memif/memif_socket.c:26:
In function 'memif_socket_create',
inlined from 'memif_socket_init' at net/memif/memif_socket.c:965:12:
net/memif/rte_eth_memif.h:35:2: error:
'%s' directive argument is null [-Werror=format-overflow=]
35 | rte_log(RTE_LOG_ ## level, memif_logtype, \
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
36 | "%s(): " fmt "\n", __func__, ##args)
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Fixes: 09c7e63a71 ("net/memif: introduce memory interface PMD")
Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Added a note to the getting started guides about patching third
party libraries/dependencies to avoid any known vulnerabilities.
Signed-off-by: John McNamara <john.mcnamara@intel.com>
As legacy filter API "filter_ctrl" is superseded since 2017
by the rte_flow API, and got the deprecated attribute in DPDK 19.05,
it is time to remove the associated features from the matrix.
Not documenting deprecated features as supported will avoid confusion.
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Andrew Rybchenko <arybchenko@solarflare.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
