This plugin implements a PoC of UDP-based tunnel substrate whose aim is
to specifically provide higher MTU to the upper layers by chunking
the payload PDUs into smaller packets with full 5-tuple.
At the same time, if there are multiple small packets to
the same destination during the vector processing, they
are packed into "carrier" packets up to underlay MTU size.
It does assume a trustworthy underlying medium, thus for the
operation over Internet it requires the use of encryption layer
underneath.
Type: feature
Change-Id: I323958fa8de62584f6ed15643ea689568a9a62bc
Signed-off-by: Andrew Yourtchenko <ayourtch@gmail.com>
This patch bumps DPDK version to 24.07 and RDMA version
per dpdk 24.07 release notes (rdma-core >= 52.0)
Type: feature
Change-Id: I07c6f27becdbcd3696b821bbfd145610775cf51e
Signed-off-by: Kai Ji <kai.ji@intel.com>
Add support for:
- "sh session verbose thread <n> [lcl|rmt|ep] <ip>[:port]", which can be
used to filter sessions based on transport ip:port
- "force-print" option to force printing even when it exceeds existing
printing thresholds
Type: improvement
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: Iab3e58b8162664b425379d14818b8a4daa0ce345
Make sure that proxy server doesn't try to open connection to the
target multiple times. When client is uploading huge amount of data
it might happen that proxy_rx_callback is called again before
connection to the target is established.
Type: fix
Change-Id: I4d272d37bc545f6f8790a115c8d26bcf3ea84f22
Signed-off-by: Matus Fabian <matfabia@cisco.com>
The endian function for VLAs assumed length field to be of type u32.
That failed for APIs using different integer width.
Type: fix
Change-Id: I6ecaabb4563c8bafeb100a4c6c1eee9a08e6cabf
Signed-off-by: Ole Troan <otroan@employees.org>
This change allow a node registration A to name a next node B that does not
exist yet at registration time. When node B is effectively created,
vlib_node_main_lazy_next_update need to be called so that the vlib graph
is updated accordingly. To enable this behavior, node A needs to bear
the new VLIB_NODE_FLAG_ALLOW_LAZY_NEXT_NODES.
Change-Id: I561d3a0de19a0b7bd1045760a2ba8e27d27caa9a
Type: improvement
Signed-off-by: Mohammed Hawari <mohammed@hawari.fr>
Reuse existing proxy_force_ack for active open tx callback. This makes
sure proxy session is tcp (as opposed to tls) and also makes sure the
ack is sent from the thread that owns the proxy session.
Type: fix
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: I9692fa2c4feefd3db701f9449a35495cf7b3d9dc
This patch removes redundant sa index value in
format_ipsec_tun_protect_input_trace() as sa index value already added
by format_ipsec4_tunnel_kv().
Type: fix
Change-Id: I7834bda57ba0f7cfe5084b671f6f7da0749ba64d
Signed-off-by: Garvit <gvarshney@marvell.com>
With this feature, session enable is now modified to have 3 modes of operation
session enable -- only enable session
session enable rt-backend sdl -- enable session with sdl
session enable rt-backend rule-table -- enable session with rule-table
session rule tables are now created on demand, upon adding first rule
to the rule table.
refactor session table to remove depenency from sesssion rules table. Now
session rules table APIs take srtg_handle and transport
proto instead of srt pointer.
Type: feature
Change-Id: Idde6a9b2f46b29bb931f9039636562575572aa14
Signed-off-by: Steven Luong <sluong@cisco.com>
Now when lcp pair is created, tap instance is based on hw_id. But tap
interface with such instance can already exist. Introduce an offset and
auto-selection based on it.
Type: fix
Signed-off-by: Stanislav Zaikin <stanislav.zaikin@46labs.com>
Change-Id: I9db39106b0b0d5bf95c445b03e5b7ff52f946dd2
Type: fix
The flowprobe-input-ip4 node must be run ahead of ip4-inacl in some scenarios
to ensure that it sees flows on receiving interfaces.
The same is presumably the case for flowprobe-input-ip6, but this hasn't been
tested/observed.
JIRA: VPP-2122
Change-Id: Idf3b14dd2bd829f9aa3b1a71ccdcdf015ccdb36a
Signed-off-by: Scott Hutton <schutton@cisco.com>
In some multi-process/thread applications, a connection FD
is created on main process and actual data processing is done
in worker process. The main process does nothing. For such
applications, when multi-thread-workers is enabled, in current
VLS code, the session events are still enqueued to the VCL
worker corresponding to Main process. As the main process is
not dequeuing any event, application doesn't move forward.
This patch fixes this issue by enabling VCL worker
corresponding to the Worker process as listener of the session.
Type: improvement
Signed-off-by: Mohamed Feroz <mabdulmajeet@marvell.com>
Change-Id: Ia7e6270c1acbce7feeafbf281c661285d63c3b22
Previously, iavf_port_add_del_eth_addr was not using large enough buffer
and address sanitizer does not allow that.
Type: fix
Change-Id: Icd1491fb5651aed20685d15224e9c725347ef369
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
Add RSS support to make use of multiple queues.
With 4 RX queues and RSS enabled
ping from host to guest vm ip queue use
192.168.2.3 192.168.2.1 2
'' 192.168.2.10 0
'' 192.168.2.5 1
'' 192.168.2.105 3
With 4 RX queues and RSS disabled, queue 0 is always used for all of the above cases
Type: improvement
Change-Id: I3ca78fd83fce26cbe8f23fee0a9034cb572bacb7
Signed-off-by: Steven Luong <sluong@cisco.com>
Currently, the per-numa buffer allocation doesn't support skipping the
allocation altogether for a particular NUMA domain - if it's set to
zero, it instead uses the default allocation.
Therefore, this changes the behavior so that the default is only used
if no per-numa allocation is specified, and an allocation of zero
is honored by not allocating a buffer at all.
Type: improvement
Change-Id: Ie9133b179a88eb884aa3da921e65292daea0df2d
Signed-off-by: Lukas Stockner <lstockner@genesiscloud.com>
vnet flow enable is failing due to bogus bytes pointed by spec, mask
variables of roc_npc_flow_item structure. Using reference to local
variables defined in block scope is causing this. Moving the variable
declarations to function block scope fixes this issue.
Fixes: 064762e20
Type: fix
Signed-off-by: Sriram Vatala <svatala@marvell.com>
Change-Id: I3904199b5b2bd88cd02ada5604059ab6fd12eef7
On FreeBSD the affinity returned by cpuset gives us a bitmask
documenting the available cpus. This can be modified if we are jailed or
are launched via the cpuset(1) command.
Initialise the clib bitmask otherwise it is always 0 and we cannot
configure any workers.
Type: improvement
Change-Id: I067e373ea440c6fbd03839d1cf103c8b97d8a0c6
Signed-off-by: Tom Jones <thj@freebsd.org>
set http status according to whether app also sent the body
Type: improvement
Change-Id: Ia41603cc21b410ca6929ec3d3e7c4c6808305769
Signed-off-by: Matus Fabian <matfabia@cisco.com>
- basically a copy of LdpIperf test
- small update of LdpIperf test
- new LDP suite
Type: test
Change-Id: I3f8653288c6fc6dfd6a061315e983c000974d3ff
Signed-off-by: Adrian Villin <avillin@cisco.com>
Remove failed-summary.log from the previous run, otherwise it might be
displayed if all tests pass.
Type: test
Change-Id: I435de37e298105ed5b92012fe4da3cda3a0670b6
Signed-off-by: Matus Fabian <matfabia@cisco.com>
Request line must only start with method name and server should
ignore at least one empty line (CRLF) received prior to the
request-line.
Type: fix
Change-Id: Ifebd992dc4c13df1a3fabfcdef9e7ee644150a21
Signed-off-by: Matus Fabian <matfabia@cisco.com>
- load vpp plugin libraries path in gdb
- if core dump deceted change message in summary and show bt
Type: test
Change-Id: If5d44d29703edd16c816efc9d29e6a7a1b20613a
Signed-off-by: Matus Fabian <matfabia@cisco.com>
Use ExpectWithOffset so Ginkgo's output point to the line in your test.
Type: test
Change-Id: Icb32c5be683f935a5e86f25c2ec5a6dfad2ec1ea
Signed-off-by: Matus Fabian <matfabia@cisco.com>
- exit status was incorrect when tests didn't compile
- better summary formatting
- removed "full stack trace: null" from summary when a test times out
Type: test
Change-Id: Idd6b6bf3429fc8b704dd2345d257e9bdce1b82b7
Signed-off-by: Adrian Villin <avillin@cisco.com>
Type: refactor
This patch moves osi into a plugin, and also modifies
the init functions of llc and snap to preserve init
order dependency (llc_init --> osi_init --> snap_init).
While the initial intent was to move osi/llc/snap together
into a single plugin, there exists a dependency on llc
in vnet/ethernet, which would require further refactoring
and testing work.
Change-Id: Ic0eff030ee29c8d316c0e0fe13931451aa193527
Signed-off-by: Hadi Rayan Al-Sandid <halsandi@cisco.com>
- add patch to scapy to use latest six module instead
of the old scapy specific six module which fails to
import moves sub-module on python 3.12
- fix warning for deprecated legacy editable install of vpp_papi
- skip failing testcases on ubuntu-24.04
Type: test
Change-Id: Idc1a008c6e45ba69caa50c2e245012eb05effed7
Signed-off-by: Dave Wallace <dwallacelf@gmail.com>
- test moved to proxy_test.go
- topology and suite updated to be vpp/envoy proxy compliant
Type: test
Change-Id: Iabee50b76bb8d96851f831c0b57d0e98dc4b3a5f
Signed-off-by: Matus Fabian <matfabia@cisco.com>
