if the api client didn't wait for the last message, we'd get a SIGPIPE
from Unix and VPP would crash.
Type: fix
Signed-off-by: Ole Troan <ot@cisco.com>
Change-Id: Iac7705ec09ccd67cc249cc9a9525a7cb379e2f6f
Signed-off-by: Ole Troan <ot@cisco.com>
Async use of the API is much faster than blocking calls.
Seemed like it only worked over shared memory transport.
This patches re-enables support for async calls over socket
transport.
Type: fix
Signed-off-by: Ole Troan <ot@cisco.com>
Change-Id: I05f3b362035ce0a1c16788ba9003a35601ddb04e
Signed-off-by: Ole Troan <ot@cisco.com>
The request vector generated by hss_ts_rx_callback() must be NULL
terminated.
The hss_main_t use_ptr_thresh member must be a u64 since
unformat_memory_size() expects it. Otherwise, the adjacent u8
enable_url_handlers may have an accident.
Type: fix
Signed-off-by: Dave Barach <dave@barachs.net>
Change-Id: I2cc08e3cbd31b225fb03799283c055515add13bf
The metadata in classifier entries is used to index a fib or a dpo in
the acl nodes which can exceeds UINT16_MAX in large configurations.
To maintain entries size and alignment, decrease next_index from 32- to
16-bits: next_index should not exceed 16-bits in VPP, as it is already
shown by vlib_buffer_enqueue_to_next() or dpo_id_t.dpoi_next_node.
Type: fix
Change-Id: I4fd1b3cd495319420044c219036b2d2ea952270a
Signed-off-by: Benoît Ganne <bganne@cisco.com>
This adds basic, functional-only, test of TCP connection with delay and
packet loss introduced by Network Delay Simulator.
Type: test
Signed-off-by: Maros Ondrejicka <maros.ondrejicka@pantheon.tech>
Change-Id: Ibedf4c680c152921b733cf39d99b178412748d3c
Add an http redirect template to generate correct-looking "301 Moved
Permanently" replies.
Supply a default value of 1<<31 for the use_ptr_thresh config parameter.
Expose hss_session_get() so friend plugins which register GET / POST
handlers with the http_static server can add data to the session fifos.
Type: fix
Signed-off-by: Dave Barach <dave@barachs.net>
Change-Id: Ie1452eaf61c6f67311fbab092bc1fe03050bf94f
The return value in this function is initialized with 0, so if a vector of length 0 is passed
to stat_segment_dump_r, then this return value is never populated, resulting in inability
to distinguish between a successful dump of an empty vector and an error.
Solution: call vec_alloc(). As a side effect might get some trivial speed-up.
Type: fix
Signed-off-by: Andrew Yourtchenko <ayourtch@gmail.com>
Change-Id: I33fefd801df457152e9ec257742305182e91f339
Test suite now supports assertions which on fail stop test case run,
also it allows to create docker containers which are going to be
stopped automatically after the test run is finished.
Type: improvement
Signed-off-by: Maros Ondrejicka <maros.ondrejicka@pantheon.tech>
Change-Id: I2834709b1efd17b8182d36cc0404b986b4ed595d
Signed-off-by: Filip Tehlar <ftehlar@cisco.com>
This change was introduced to workaround a bug in the NAT code, but
we should not woraround plugin bugs in infra.
Type: fix
Fixes: f8631ce7e8886136b4543a7926ffdf1bc760fb11
Change-Id: Id6ee281cf1fe8466b6522905fc2a176716e3d52f
Signed-off-by: Benoît Ganne <bganne@cisco.com>
Account for the potential of sysconf() returning -1 if it can not
get the page size and make it a fatal error.
Coverity: 277313
Type: fix
Signed-off-by: Andrew Yourtchenko <ayourtch@gmail.com>
Change-Id: I8cae6a35ec2f745c37f1fe6557e5fa66720b4628
convert vhost device driver to a plugin as described in
https://jira.fd.io/browse/VPP-2065
Type: improvement
Signed-off-by: Steven Luong <sluong@cisco.com>
Change-Id: Ibfe2f351bcaed36a04b136d082ae414145dd37b5
Support generic flow in native avf.
Enable necessary RSS hash function for generic flow. Extend some
structures and functions from for FDIR only to for both RSS and FDIR
flows. Modify virtual channel message to align with ice kernel driver.
Add functions to parse generic flow patterns. The parsing results will
be delivered to the kernel driver and create corresponding flow rules.
Type: feature
Signed-off-by: Ting Xu <ting.xu@intel.com>
Change-Id: I82ce102a21993f1bae8a8bf23e491d5e1c261f61
The initiate handshake process can be called a numbers times for each
peers, then the main VPP thread called by Wireguard starting to
allocate memory. This behaviour can lead to out of memory when VPP has
a lot of Wireguard tunnels concurrently.
This fix add mutex to send only once handshake initiate at time for
each peers.
Type: fix
Signed-off-by: Gabriel Oginski <gabrielx.oginski@intel.com>
Change-Id: I13b4b2d47021753926d42a38ccadb36a411c5b79
Write time into /sys/boottime on VPP start.
This allows a stateless control plane agent to validate if it's reconnecting to the same
VPP instance.
Type: improvement
Signed-off-by: Ole Troan <ot@cisco.com>
Change-Id: Iba7f334339c46142045e43da6efab11612e7b9c0
Signed-off-by: Ole Troan <ot@cisco.com>
Maintain a single writer multiple readers usage model for transport
endpoints pool.
Type: improvement
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: I8555700ed725971341f145ea97f031042a298e83
The issue can be reproduced by running "vpp_get_stats tightpoll"
The root cause is that the control flow discards the "result" struct
being prepared, along with pointer its allocated name.
This results in a memory leak.
Type: fix
Change-Id: Ibf884e92314f19b983a0159fc1257b3fa0110443
Signed-off-by: Andrew Yourtchenko <ayourtch@gmail.com>
When n_rx_packets is less then 16(VEC256) or 8(VEC128), code execution
will fall into scalar path of processing packets. But with a wrong
initialization value for n_left set to zero, i in the for-loop will
equal to n_rx_packets. This leads to the bypass of required ip4 checksum
validation and byte count endianness conversion in scalar path.
Besides, refactor the code using while instead of for-loop to keep
consistency with VPP code style.
Type: fix
Fixes: bf93670c515d ("rdma: fix ipv4 checksum check in rdma-input node")
Signed-off-by: Lijian Zhang <lijian.zhang@arm.com>
Signed-off-by: Jieqiang Wang <jieqiang.wang@arm.com>
Change-Id: Ib4e8cb5202735f8b060c99caddf26035657551e1
Type: fix
Fixes: a51f9b3747
Some IPsec message type definitions were not being written to
ipsec.api.vapi.h. These include ipsec_sad_entry_add_del_v3 and
ipsec_sad_entry_add.
The cause appears to be that tunnel_flags, which is defined in
tunnel_types.api is a special case of enum called an enumflag. These do
not appear to have been handled in the code that generates the vapi
header files.
This patch adds processing of enumflag objects for vapi.
Change-Id: Ie506c4fcb5a07fe97a330ba11c252d1df98adfd9
Signed-off-by: Matthew Smith <mgsmith@netgate.com>
Fast path spd was explicitely storing array of policy id vectors.
This information was redundand, as this inofrmation is already stored
in bihash table. This additional array was affecting performance
when adding and removing fast path policies.
The other place that needed refactoring after removing this array was
cli command showing fast path policies.
Type: feature
Signed-off-by: Piotr Bronowski <piotrx.bronowski@intel.com>
Change-Id: I78d45653f71539e7ba90ff5d2834451f83ead4be
Not ideal. The sparse vector used to map ports to next nodes assumes
only a few ports are ever used. When udp transport is enabled this does
not hold and, to make matters worse, ports are consumed in a random
order.
This can lead to a lot of slow updates to internal data structures
which in turn can slow udp connection allocations until all ports are
eventually consumed.
Consequently, reallocate sparse vector, preallocate all ports and have
them point to UDP_NO_NODE_SET. We could consider switching the sparse
vector to a preallocated vector but that would increase memory
consumption for vpp deployments that do not rely on host stack.
For reference, populating one of the v4 or v6 sparse vectors in reverse
order takes about 9.8s on a skylake cpu.
Type: improvement
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: Id795e1805d0d3ba54f56a152a9506a7a2a06ecbc
Implements the API for SRv6 Path Tracing
Type: feature
Signed-off-by: Julian Klaiber <julian@klaiber.me>
Change-Id: Iefa7e512c8e1894595a9e3f5d42eab4160db1f28
In order to loop over the list of `vl_api_ip_address_details_t`,
increment the pointer by one instead of `i`.
Type: fix
Change-Id: I8554d1388d67bb95e029eddf444d383fb85ecac7
Signed-off-by: Timur Celik <mail@timurcelik.de>
Type: fix
Fib entries for attached routes when sourced from
FIB_SOURCE_API or FIB_SOURCE_CLI
get the FIB_ENTRY_FLAG_ATTACHED flag raised on the source.
Such a route added from linux-cp doesn't get this flag.
Fix this flag for linux-cp sources by passing it to the
fib entry's update/create function in lcp_router_route_add().
Signed-off-by: Alexander Skorichenko <askorichenko@netgate.com>
Change-Id: I24278ef86886cfee8a14acb250fb6992a754cc3c
Type: docs
Several kinds of policers are implemented in VPP.
However, they could differ from the RFCs it is
said they are from.
Additionally, the CLI command's help has been
updated with the current list of acceptable
parameters.
Signed-off-by: Maxime Peim <mpeim@cisco.com>
Change-Id: Ic9bf94e1094bea0fcc87ccaa882c2c5f88824041
We add the crypto key to the vnet crypto library via vnet_crypto_key_add.
However, when the session is disconnected, we don't call
vnet_crypto_key_del and the memory is leaked in vnet_crypto library
as well as in pico tls key store.
It seems dispose crypto is the appropriate place to add
vnet_crypto_key_del.
Type: fix
Signed-off-by: Steven Luong <sluong@cisco.com>
Change-Id: If6d1266baf686fefe5bb81330ce60b35c8ff574e
Plugin checks just for AVX2 instruction set, while the v1.3 of IPsec
Multi-Buffer library checks for both AVX2 and BMI2 sets during init.
VirtualBox VM doesn't provide BMI2 by default to guest operating system.
Result is that VPP plugin decides to use AVX2 initialization and library
then doesn't do it. Since flush_job remains empty, the self-check fails
and with that the whole VPP crashes on start-up.
Type: fix
Signed-off-by: Maros Ondrejicka <maros.ondrejicka@pantheon.tech>
Change-Id: I6b661f2b9bbe6dd03b499c55c38a9b814e6d718a
Make sure they only double in size.
Type: improvement
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: I18d5508c7f32836deb3b25943e8e3af39d0dbc33
