session table may be shared among multiple appns's.
app ns add id blue secret 1 if tap0
app ns add id red secret 1 if tap0
session table holds the last added app_ns's appns_index.
If the last app_ns is deleted, session table is not free
since there is still an appns which uses the same session
table. In that case, session table is holding the free
app_ns's appns_index and it can cause problem.
The fix is to modify appns_index in session table to hold
a vector of appns_index's instead of just the appns_index
that was last added. When the app ns is deleted, remove the
deleted appns_index from the session table's vector of
appns_index's.
Type: fix
Change-Id: Ied8bc97f185071dc89b9b56656e18efbd2995131
Signed-off-by: Steven Luong <sluong@cisco.com>
Make sure num threads is 1 on process create and fork. Multi-thread
locks are applied once num threads exceeds 1.
For select, follow same pattern like epoll and add check for session
migration.
Type: fix
Signed-off-by: Florin Coras <fcoras@cisco.com>
Signed-off-by: Dave Wallace <dwallacelf@gmail.com>
Change-Id: I1edcd6c4c81b6b3caf8b00781b339414e8945b0e
In certain usecases related to Linux legacy pinning of flows on queue
pairs, it is desirable that, for a given index, the rx and tx virtio
queue be handled by the same worker. This change introduces a flag for
virtio and tap interfaces that allow such a mapping.
Example with two workers rxq 0 and txq 0 on worker 0
rxq 1 and txq 1 on worker 1
txq 2 on main thread
Change-Id: I1b74a4788843fd1d0e8dcb4e9da30e609e088fe3
Signed-off-by: Mohammed Hawari <mohammed@hawari.fr>
Type: improvement
The convention in the binary API is that fields encoded in network order.
For some reason, port was parsed in host order.
Type: fix
Change-Id: I31ea313937097e2547226566b7869be4e28251b8
Signed-off-by: Steven Luong <sluong@cisco.com>
Parse a URI template that has variables "target_host" and "target_port",
where varaibles are at the end of the path:
"/{target_host}/{target_port}/".
Type: improvement
Change-Id: I440b7f4951bffa1fd9971740b9890b221193943b
Signed-off-by: Matus Fabian <matfabia@cisco.com>
This patch fixes handle with UDP packages for UDP-Encapsulated ESP
and IKE traffic in inbound policy.
Orignally IKE traffic on UDP with port 4500 is dropped inside inbound
look-up.
Type: fix
Change-Id: I071adf18cb82da8cd000b93914078e51e393104c
Signed-off-by: Gabriel Oginski <gabrielx.oginski@intel.com>
Treat all connects as connected udp sessions. Connectionless udp is only
possible with listeners.
Type: improvement
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: I0a75b7c85b72274df909ed3b114eed4f6519477e
- mark listener as connected to support udp proxying. Connectionless udp
proxying is possible but would need more work on rx path.
- add support for session migration
Type: improvement
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: Ia6911ad728b7718a4983152d144e2566998e35c2
App can now pass http connection timeout using extended configuration,
ext cfg type TRANSPORT_ENDPT_EXT_CFG_HTTP, value (in seconds) set in ext
cfg member opaque.
It is optional, default value is 60 seconds.
Type: improvement
Change-Id: Ibeff4bbd3153421be350ff564ec3c8e52e5b9639
Signed-off-by: Matus Fabian <matfabia@cisco.com>
This allow configuration for each transport protocol when combined,
e.g. HTTPS=HTTP+TLS.
Extended config of session endpoint config is now chunk of memory,
which allow seamless integration with VCL, and internal
representation is hidden behind APIs, which is better for
future improvements. By default chunk is allocated to 512B if the
app doesn't do so before.
Type: improvement
Change-Id: I323f19ec255eba31c58c06b8b83af45aab7f5bb1
Signed-off-by: Matus Fabian <matfabia@cisco.com>
Previously the DHCP client use a constant client identifier (vpp 1.1).
Problem with that is that multiple VPP clients on the same link would
fail. Fixed by using MAC address instead.
Type: improvement
Change-Id: If37f4a14cc728bde6c9ef6c169957562d763c973
Signed-off-by: Ole Troan <otroan@employees.org>
When the last listener is deleted, detach the cli server.
The appns is determined and should be entered when the cli
server is first brought up. When adding an additional listener
subsequently, appns must not be specified.
Type: improvement
Change-Id: Idac5b1f89df2aeca6734ab4532ebade0659fc357
Signed-off-by: Steven Luong <sluong@cisco.com>
Replace a hard-coded constant of 64 with 56, which is a more appropriate
value for when ICMP echo request is returned in ICMP error. Previously,
the size of such message would be smaller than 64 and parsing would not
return the correct value.
Type: fix
Change-Id: I1bdf8a2a23b1e6762b64210c71927c926f4547ad
Signed-off-by: Klement Sekera <klement.sekera@gmail.com>
The size of dhcp option should be o.length + 2 additional bytes of the header.
Incorrect offset results in an extra byte at the end of the packet:
...
0120 04 00 00 00 04 05 04 AC 10 04 01 FF 00
...
RFC2131 says the last should be the 'end' option (ff)
Type: fix
Change-Id: I056d755d29465aab8c1c55a0b930f65ece6fafce
Signed-off-by: Artem Glazychev <glazychev@mts.ru>
Fix typo and put it on the correct level.
Type: fix
Change-Id: Ic7a8e557b22d0060d7f072132f00564e29d8e860
Signed-off-by: Lukas Vogel <vogel@anapaya.net>
The default number of tx descriptors per queue is 1024. On some device
types, this is larger than the maximum allowed value and
rte_eth_tx_queue_setup() can return an error. Compare the configured
value to the maximum value for each device and set the configured
value to the maximum value if the configured value is larger.
Type: improvement
Signed-off-by: Matthew Smith <mgsmith@netgate.com>
Change-Id: Ie29c5c6d58c76388f65f266032821140b0f879fb
Type: test
Activate tests that works successfully on ubuntu 22.04 but were
problematic on later version and were excluded by
@tag_fixme_ubuntu2204
Change-Id: I93c4626a120ce8b475e9c5ca933a5544d6337933
Signed-off-by: Ivan Ivanets <iivanets@cisco.com>
