Age | Commit message (Collapse) | Author | Files | Lines |
|
Type: refactor
Change-Id: I5235bf3e9aff58af6ba2c14e8c6529c4fc9ec86c
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
When link state interval is 3 and stats interval is 10, updates
for stats will be made every 12 seconds (next number after 10,
that divisible by 3). And if you get counters every 30 secs, you
will get ideal "saw"-line instead of real smooth chart.
This commit makes smooth line on stats intervals that are divisors
of the charts update interval (regardless of link state interval),
and makes it possible to configure them.
Type: fix
Fixes: cb9cadad578297ffd78fa8a33670bdf1ab669e7e
Change-Id: Ia4350467be2b0ec0c1be37c7fda63f43b3330f44
Signed-off-by: Georgy Borodin <bor1-go@yandex-team.ru>
|
|
As similar 535364e90459566b603661c3dbe360c72f59ad71 is
merged, printing possibly deleted interfaces by index
only in all the rest cases.
Type: improvement
Signed-off-by: Vladislav Grishenko <themiron@yandex-team.ru>
Change-Id: I4fa58b382c0279ff893523ba0188fdb9b09e10af
|
|
Type: feature
Enable use of 4th gen QAT series devices.
Change-Id: I890c1f1d305ff9b996322c29e9510cfe89d88d97
Signed-off-by: Kai Ji <kai.ji@intel.com>
|
|
debug+asan build will fail on initialisation when log level==debug
Type: fix
Fixes: 549838c81bd0d995f2b8569955afc33132582c77
Change-Id: Ic64b376727d49f89790fb29bd03a0a2520625cdc
Signed-off-by: Georgy Borodin <bor1-go@yandex-team.ru>
|
|
Even if the device name is specified in the startup config it may
be appended to if the device is a switch domain member. This leads
to unexpected device naming if an explicit device name was requested.
Type: fix
Change-Id: Ib56b4ac41c17008db55dc69497721e3cb7d540c1
Signed-off-by: Peter Morrow <pdmorrow@gmail.com>
|
|
This patch introduces a fix for correcting a counter for the number
of processed vectors in the crypto-dispatch node.
Type: fix
Signed-off-by: Piotr Bronowski <piotrx.bronowski@intel.com>
Change-Id: Icaeb925a352a9ac766652f43c4e752f6727cdeb9
|
|
This is because mlx5_pci is also compatible with another
series of NICs such as ConnectX-5 and ConnectX-6.
Type: fix
Change-Id: I10f0468bbe36ab61c72fb3dc0aa898f8e2f9e88c
Signed-off-by: Nobuhiro MIKI <nmiki@yahoo-corp.jp>
|
|
DPDK added new Rx checksum flags[1] to handle cases like the virtual
drivers. Current check of flags is not strict enough for flags like
RTE_MBUF_F_RX_IP_CKSUM_NONE and will always be true no matter the
checksum in packet is good or bad.
Fix this issue by comparing the result of AND operation with the
correspinding Rx checksum flags.
Before this patch, packet trace prints the offload flags as below:
Packet Offload Flags
PKT_RX_IP_CKSUM_GOOD (0x0080) IP cksum of RX pkt. is valid
PKT_RX_IP_CKSUM_NONE (0x0090) no IP cksum of RX pkt.
PKT_RX_L4_CKSUM_GOOD (0x0100) L4 cksum of RX pkt. is valid
PKT_RX_L4_CKSUM_NONE (0x0108) no L4 cksum of RX pkt.
After this patch, packet offload flags would be like:
Packet Offload Flags
PKT_RX_IP_CKSUM_GOOD (0x0080) IP cksum of RX pkt. is valid
PKT_RX_L4_CKSUM_GOOD (0x0100) L4 cksum of RX pkt. is valid
Type: fix
[1] https://github.com/DPDK/dpdk/commit/5842289a546ceb0072bd7faccb93821e21848e07
Signed-off-by: Jieqiang Wang <jieqiang.wang@arm.com>
Change-Id: I3182022d9ccd46b2fc55bb3edfbfac9062ed7c89
|
|
Type: improvement
Change-Id: Ie86a5edf2ada21355543e9a0382052b16ff86927
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Type: improvement
Change-Id: If775b1d145e462346de562a3c893f302e8c7b814
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
List BlueField NICs as a supported PCI devices.
Type: feature
Change-Id: Ida2300df516ab9cd2fcde1f816bbdc081016039a
Signed-off-by: Alexander Kozyrev <akozyrev@nvidia.com>
|
|
This patch adds support for using l2tpv3 as RSS type
Type: feature
Signed-off-by: Xinyao Cai <xinyao.cai@intel.com>
Change-Id: Ic3e0935a4754d084184f1cc38ea9531ddfd9e7bc
|
|
This patch provides minor improvements to the logic governing dequeuing
from the ring. Previously whenever a frame was dequeued
we've been trying to dequeue from the ring another one till
inflight == 0. Now threshold is set for 8 frames pending in the cache
to be consumed by the vnet. This threshold has been chosen based on
cache ring stats observation in the system under load.
Some unnecessary logic for setting deq_tail has been removed.
Also logging has been corrected, and cache ring logic simplied.
Type: improvement
Signed-off-by: Piotr Bronowski <piotrx.bronowski@intel.com>
Change-Id: I19f3daf5913006e9cb23e142a163f596e85f5bda
|
|
List Mellanox ConnectX-6LX and ConnectX-7 as a supported PCI devices.
Type: feature
Change-Id: Ieeca3f214d08f29238c387354055ac1320cab75f
Signed-off-by: Alexander Kozyrev <akozyrev@nvidia.com>
|
|
The logic for calcuating processed elements in the cache ring was broken.
In case tail and deq_tail equals and frame element pointed by the tile
is not NULL it means there is exactly one processed element in the ring.
Type: fix
Signed-off-by: Piotr Bronowski <piotrx.bronowski@intel.com>
Change-Id: I69c978334fc952049393214ccc9cc5245351f7f7
|
|
This patch addresses coverity issues CID 322716 and CID 322717.
Type: fix
Signed-off-by: Piotr Bronowski <piotrx.bronowski@intel.com>
Change-Id: I59d6f40c1af8e829d8cb3c042a52e144aeaf1e6b
|
|
Sw ring is renamed to the cache ring. This name better reflects the
puropse of this ring. We've introduced push/pop functions, as well as
other utility functions which remove code repetition. Error handlig
is improved: previously in case of an error all frame elements were
marked as bad, now only these for which errors occured have the error
status set.
Unnecessary stats counters have been removed.
Type: improvement
Signed-off-by: Piotr Bronowski <piotrx.bronowski@intel.com>
Change-Id: I2fd42a529ac84ce5ad260611d6b35a861d441c79
|
|
clang-16 complains about signed one-bit bitfield value changes
from 1 to -1. Use unsigned type instead.
Type: fix
Signed-off-by: Tianyu Li <tianyu.li@arm.com>
Change-Id: I84f8cf314d36183a5e6f544cd756c01d1d10a1a5
|
|
This patch introduces sw_ring to the crypto op data path implementation,
so that raw data path and crypto op data path use same mechanism of processing
async frames. Crypto op ring has been removed from the implementation.
Type: improvement
Signed-off-by: Piotr Bronowski <piotrx.bronowski@intel.com>
Change-Id: Id823f80a88cfa0ff40252616a36de8bb044c7f45
|
|
When vlib buffer is processed on vnet side its length is corrected by
cipher padding and icv_sz. These changes need to be reflected in
the mbuf internals.
Type: fix
Signed-off-by: Piotr Bronowski <piotrx.bronowski@intel.com>
Change-Id: I0aa03f67f556dfc8f9a577ca1967210527221e02
|
|
This rework tries to address issues found on SPR QAT, for traffic
reaching max possible throughoutput for single QAT PF packet drops were
observed.
Fix changes enq/deq scheme by utilizing software ring in enq call from
VNET but enq and deq to QAT happens only in deq callback function what
should enable better utlization of hardware resources.
Type: improvement
Signed-off-by: Dastin Wilski <dastin.wilski@gmail.com>
Signed-off-by: Piotr Bronowski <piotrx.bronowski@intel.com>
Change-Id: I2e8c473d20a269fd5e93f0c8d1f8c8aa193712bd
|
|
This patch introduces sw_ring. This ring is used in next set of patchas
and plays role of a buffer for QAT, allowing collecting frame elements
in case QAT queue is fully utilized, and assembling frame
from QAT dequeued elements.
Type: improvement
Signed-off-by: Piotr Bronowski <piotrx.bronowski@intel.com>
Signed-off-by: Dastin Wilski <dastin.wilski@gmail.com>
Change-Id: I20718e200986ab4dba5cbc31c05a904072a6981a
|
|
This patch can make crypto dispatch node adaptively switching
between pooling and interrupt mode, and improve vpp overall
performance.
Type: improvement
Signed-off-by: Xiaoming Jiang <jiangxiaoming@outlook.com>
Change-Id: I845ed1d29ba9f3c507ea95a337f6dca7f8d6e24e
|
|
Signed-off-by: Andrew Ying <hi@andrewying.com>
Type: fix
Change-Id: I3c428c90146387ad9ce291c7f646d74f06952b40
|
|
When I setup vpp by netvsc driver, occurs the following crash:
(format_dpdk_device_name) assertion `(i) < vec_len (dm->devices)' fails
vnet[100166]: #6 0x00007f434d651f6a _clib_error + 0x2da
vnet[100166]: #7 0x00007f430b4bef64 format_dpdk_device_name + 0xf4
vnet[100166]: #8 0x00007f434d6555f3 do_percent + 0xee3
vnet[100166]: #9 0x00007f434d654359 va_format + 0xb9
vnet[100166]: #10 0x00007f434d7ac16e vlib_log + 0x3ce
vnet[100166]: #11 0x00007f430b49ebe3 dpdk_device_start + 0x193
vnet[100166]: #12 0x00007f430b4aa233 dpdk_interface_admin_up_down + 0x163
vnet[100166]: #13 0x00007f434d988fc8 vnet_sw_interface_set_flags_helper + 0x378
vnet[100166]: #14 0x00007f434d989338 vnet_sw_interface_set_flags + 0x48
This patch fix it by device_index as a index for devices vec, and not
dpdk port_id.
Type: fix
Change-Id: I84c46616d06117c9ae3b2c7d0473050f1b8ded5f
Signed-off-by: Daniel Ding <danieldin95@163.com>
|
|
Type: fix
Signed-off-by: Xiaoming Jiang <jiangxiaoming@outlook.com>
Change-Id: I9971e69135e0652a36e4b4754774a43ea1d92e8b
|
|
Prior to dpdk-22.11, VPP can count on rte_eth_dev_socket_id to return
numa node 0 if the device didn't set it. Ever since below patch is
committed in dpdk
https://patchwork.dpdk.org/project/dpdk/patch/20220929120512.480-1-olivier.matz@6wind.com/#152498
the aforementioned assumption is no longer true. If the device didn't
set the numa node, VPP gets -1 from the aforementioned API call. This
causes VPP to crash.
This fix is to set the numa node to 0 if the API returns -1, or SOCKET_ID_ANY
Type: fix
Change-Id: I2fde2870e5a3eb98473fe8d119fef594bfba9a8d
Signed-off-by: Steven Luong <sluong@cisco.com>
|
|
This patch prepares code for bumping DPDK version to 22.11, but the DPDK version of this patch keeps at 22.07 for compatibility.
the "no-dsa" parameter in DPDK configuration is removed, the "blacklist" parameter can be used to block the related DSA devices.
Type: feature
Signed-off-by: Xinyao Cai <xinyao.cai@intel.com>
Change-Id: I08787c6584bba66383fc0a784963f33171196910
|
|
Originally the name for each session pool is incorrectly prepared.
It doesn't have right length. It is not null terminated.
The fix corrects the name formatting for each session pool.
Type: fix
Signed-off-by: Gabriel Oginski <gabrielx.oginski@intel.com>
Change-Id: I67da3d64702ccb27a5907825528f8c95d91040bb
|
|
Originally the name for each session pool can be incorrect prepared.
The fix changes formatting for name for each session pool.
Type: fix
Signed-off-by: Gabriel Oginski <gabrielx.oginski@intel.com>
Change-Id: I42e0752f9f46c5a42524ec7b863a7c9dd3c23110
|
|
Recognize and drive google virtual ethernet (gve) in google cloud.
Type: feature
Signed-off-by: Marvin Liu <yong.liu@intel.com>
Change-Id: Ia559615ac059cabbca5d10bcd4049e87beaad638
|
|
Witout thread barrier, when dpdk_process_node initiating
dpdk lib, workers thread may also be initiating. Main
and workers threads may both setting error_main info,
that will cause memory ASAN issue.
Type: fix
Signed-off-by: Xiaoming Jiang <jiangxiaoming@outlook.com>
Change-Id: I87b73b310730719035d4985a2cff2e3308120ec2
|
|
This patch introduces IP in IP packet support for flow cli and dpdk plugin.
Specifically, the following IP in IP packet types are supported:
MAC-IPv4-IPv4-TCP/UDP/None,
MAC-IPv4-IPv6-TCP/UDP/None,
MAC-IPv6-IPv4-TCP/UDP/None,
MAC-IPv6-IPv6-TCP/UDP/None,
IP in IP flow rules can be created by using the following new keywords in vppctl:
in-src-ip, in-dst-ip : to provide information for inner IPv4 header
in-ip6-src-ip, in-ip6-dst-ip: to provide information for inner IPv6 header
in-proto : to specify inner transport layer protocol type (TCP or UDP)
in-src-port, in-dst-port : to provide information for inner TCP/UDP header
An example to create flow rule for MAC-IPv6-IPv6-TCP:
test flow add index 0 ip6-src-ip any ip6-dst-ip any in-ip6-src-ip any in-ip6-dst-ip any in-proto tcp in-src-port 1234 in-dst-port any rss function default
Another example to create flow rule for MAC-IPv6-IPv6:
test flow add index 0 ip6-src-ip any in-ip6-src-ip any rss function default
Type: feature
Signed-off-by: Xinyao Cai <xinyao.cai@intel.com>
Change-Id: I6a1ca36d47eb65b9cb5a4b8d874b2a7f017c35cd
|
|
In current flow creating process in native avf and dpdk-plugins, when
parsing the input arguments, it does not copy IPv6 src address correctly,
so that IPv6 src address will not be configured in any flow rule, and
any packet with the same address will not be matched.
Type: fix
Signed-off-by: Ting Xu <ting.xu@intel.com>
Change-Id: Ic957c57e3e1488b74e6281f4ed1df7fd491af35c
|
|
This patch allows to pass a tag when specifying
the dpdk `dev { }` interface configuration.
It allows a control plane generating a vpp.conf
file to retreive the resulting mapping between
dpdk interfaces & sw_if_indices in VPP without
having to change the interface name exposed
to the user.
Type: feature
Change-Id: I55907417de0083b82d4a127172816cec3459acf3
Signed-off-by: Nathan Skrzypczak <nathan.skrzypczak@gmail.com>
|
|
per https://jira.fd.io/browse/VPP-2058
Type: improvement
Signed-off-by: Steven Luong <sluong@cisco.com>
Change-Id: Ica0828de218d25ada2d0d1491e373c3b78179ac1
|
|
Type: improvement
Change-Id: Ifea4badd58f7e2b5e792d7506f6747851a08587f
Signed-off-by: Benoît Ganne <bganne@cisco.com>
|
|
In most cases we only need OpenSSL libcrypto (crypto primitives) but
not libssl (tls).
Type: improvement
Change-Id: I9dce27d23d65bf46aea2d0f8aaf417240701efcc
Signed-off-by: Benoît Ganne <bganne@cisco.com>
|
|
Previously it was linked and worker properly. While rdma build
was simplified, link was lost so all encrypted data won't pass
via Mellanox interfaces(ipsec, ipip, ssh etc) and NetVSC taps
won't created the right way.
Errors:
mlx5_common: Verbs device not found: 21a5:00:02.0
mlx5_common: Failed to initialize device context.
EAL: Requested device 21a5:00:02.0 cannot be used
Tested on Azure. Same errors appears on physical machine with
Mellanox connect adapter
Type: fix
Signed-off-by: Vladimir Ratnikov <vratnikov@netgate.com>
Change-Id: Ib68976282e0ed91c016a7318db6b5eddf5510c47
|
|
Type: feature
Signed-off-by: Vladimir Ratnikov <vratnikov@netgate.com>
Change-Id: I2fd1e321983ac5caa03aac8705dfc596985c35f7
|
|
Change-Id: I2040b560b2a00f8bd176ae6ad46035678a2b249e
Type: improvement
Signed-off-by: Mohammed Hawari <mohammed@hawari.fr>
|
|
Originally initialization cryptodev device(s) calls double request
to enabled async mode and increased ref count twice for async mode.
Due to this cannot be change any assigned async handlers to other
async crypto engine.
The fixes reduce double request to enable async mode in initialization
cryptodev device(s) and VPP can be change assigned async handlers
to other crypto engine after disabled all async feature, for example:
ipsec, wireguard.
Type: fix
Signed-off-by: Gabriel Oginski <gabrielx.oginski@intel.com>
Change-Id: If22e682c3c10de781d05c2e09b5420f75be151c3
|
|
This patch introduces DMA infrastructure into vlib. This is well known
that large amount of memory movements will drain core resource. Nowadays
more and more hardware accelerators were designed out for freeing core
from this burden. Meanwhile some restrictions still remained when
utilizing hardware accelerators, e.g. cross numa throughput will have a
significant drop compared to same node. Normally the number of hardware
accelerator instances will less than cores number, not to mention that
applications number will even beyond the number of cores. Some hardware
may support share virtual address with cores, while others are not.
Here we introduce new DMA infrastructure which can fulfill the
requirements of vpp applications like session and memif and in the
meantime dealing with hardware limitations.
Here is some design backgrounds:
Backend is the abstract of resource which allocated from DMA device
and can do some basic operations like configuration, DMA copy and
result query.
Config is the abstract of application DMA requirement. Application
need to request an unique config index from DMA infrastructure. This
unique config index is associated with backend resource. Two options
cpu fallback and barrier before last can be specified in config.
DMA transfer will be performed by CPU when backend is busy if cpu
fallback option is enabled. DMA transfer callback will be in order
if barrier before last option is enabled.
We constructs all the stuffs that DMA transfer request needed into
DMA batch. It contains the pattern of DMA descriptors and function
pointers for submission and callback. One DMA transfer request need
multiple times batch update and one time batch submission.
DMA backends will assigned to config's workers threads equally. Lock
will be used for thread-safety if same backends assigned to multiple
threads. Backend node will check all the pending requests in worker
thread and do callback with the pointer of DMA batch if transfer
completed. Application can utilize cookie in DMA batch for selves
usage.
DMA architecture:
+----------+ +----------+ +----------+ +----------+
| Config1 | | Config2 | | Config1 | | Config2 |
+----------+ +----------+ +----------+ +----------+
|| || || ||
+-------------------------+ +-------------------------+
| DMA polling thread A | | DMA polling thread B |
+-------------------------+ +-------------------------+
|| ||
+----------+ +----------+
| Backend1 | | Backend2 |
+----------+ +----------+
Type: feature
Signed-off-by: Marvin Liu <yong.liu@intel.com>
Change-Id: I1725e0c26687985aac29618c9abe4f5e0de08ebf
|
|
Type: feature
Change-Id: I0abbe925d6b9d3dd7196cd8beaf4f471beb45bd6
Signed-off-by: Benoît Ganne <bganne@cisco.com>
|
|
Build vpp with MLX DPDK PMD,
make DPDK_MLX4_PMD=y DPDK_MLX5_PMD=y DPDK_MLX5_COMMON_PMD=y build-release
With no-multi-seg in startup.conf,
Mellanox NIC init failed with following message,
rte_eth_rx_queue_setup[port:2, errno:-12]: Unknown error -12
mlx5_net: port 2 Rx queue 0: Scatter offload is not configured and
no enough mbuf space(2176) to contain the maximum RX packet length(2065)
with head-room(128)
In Mellanox NIC PMD driver, 'di.max_rx_pktlen' is returned as 65536,
and 'di.max_mtu' is returned as 65535, which makes
the driver_frame_overhead logic not suitable for Mellanox NICs.
So skip the logic code if MAX_MTU is returned as 65535.
Type: fix
Fixes: 1cd0e5dd533f ("vnet: distinguish between max_frame_size and MTU")
Signed-off-by: Tianyu Li <tianyu.li@arm.com>
Change-Id: I027b76b8d07fb453015b8eebb36d160b4bc8df9c
|
|
Type: improvement
Enable use of 4th gen QAT devices. Will be available on Sapphire Rapids.
Signed-off-by: Matthew Smith <mgsmith@netgate.com>
Change-Id: I89e7d29e10ecb4c36c700ff5e017796161ec6c5e
|
|
Type: fix
Correct vnet_hw_interface_t flags update on link state changes.
Currently incomplete set of flags is applied on each change,
only flags related to the most recent change are being set correct.
E.g. setting the link up would erase the duplex part of the flags.
Signed-off-by: Alexander Skorichenko <askorichenko@netgate.com>
Change-Id: I5b95e1c0eaea0c283b108dbf7f809682ec9064eb
|
|
Type: fix
There were several places where mbps were converted to kbps for
link_speed, but often drivers of devices set link speed to unknown
(0xFFFFFFFF) on initialization, so there was multiplication of
link_speed equal 0xFFFFFFFF(UINT32_MAX) by 1000, this provides
overflow of unsigned int, and as result link_speed was equal
4295 Gbps, but actually link_speed is unknown.
Signed-off-by: Anton Nikolaev <anikolaev@netgate.com>
Change-Id: Ib462ed6ed685654af4687041e115bfb74e640f13
|
|
When the driver's max_rx_pktlen is >= 65536, max_supported_frame_size
overflows and queue creation fails.
Change-Id: If78707cb698adf8619ec44a852dd05d570917577
Signed-off-by: Mohammed Hawari <mohammed@hawari.fr>
Type: fix
|