Age | Commit message (Collapse) | Author | Files | Lines |
|
Type: fix
Change-Id: I4bcc9c6168c18ccf5701b832c26b16d3bb24bc98
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Type: improvement
Change-Id: I6a6794dee3c9e948227e6c08769d45aa595bfe02
Signed-off-by: Matus Fabian <matfabia@cisco.com>
|
|
Type: fix
Change-Id: I41a9332bdd7a34ce30eaa5cd2378990dbe8e31f5
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
When a page fault occurs, and the "block on fault" flag is not set, the
DSA device reports the error in the completion descriptor and software
should complete the operation in software. This behavior works properly,
except the batch was not added to the list of pending batches, so the
DSA node failed to completely process the batch. The side effect is the
application callback is not called, so buffers are not freed, and the
batch is never added back to the freelist, leading to eventual buffer
exhaustion, and traffic stopping.
The fix is to make sure the batch is added to the list of pending
batches so the DSA node completes the batch. This change also removes
the line that increments the "submitted" counter as that is only used
to count batches that are submitted to hardware. A separate counter,
sw_fallback, is used to count batches that are handled in software.
Type: fix
Change-Id: Ib54e8d77ac02e5dca237bb9348440c3ddaad993b
Signed-off-by: Jeff Shaw <jeffrey.b.shaw@intel.com>
|
|
Apps that rely on epoll and use eventfds for mq signaling can
potentially sleep indefinitely if vcl detaches from vpp and no libc fd
generates events.
To avoid this, at detach time, force creation of a pair of pipes that
constantly generates mq_epfd events to force apps, including ldp, to
request vppcom epoll_waits which retry attaching to vpp.
Type: improvement
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: Ie2ac338cc10721829e6ac525ee7d6b812354e9a3
|
|
comparison
The comparison code for the bit remainder (non-zero part of prefix length modulo 8)
was incorrectly looking one byte further than it should.
Type: fix
Change-Id: Idd27d218e77eff5f368f2ba0a5cefb86ecf605f5
Signed-off-by: Andrew Yourtchenko <ayourtch@gmail.com>
|
|
SCHED nodes are new type of nodes, similar to input nodes but they are
scheduled to be run from timing wheel.
SCHED nodes work both on main and worker threads.
Typically SCHED nodes can be scheduled to be run in two ways:
- vlib_node_set_interrupt_pending() - from any thread, run ASAP
- vlib_node_schedule() - from own thread, afer sepcific interval
Type: feature
Change-Id: Id29a66532328d9b3c454e65d09495e8fb479cedf
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Type: feature
Change-Id: Ia6581b685c9dbc89e72bdec320b7f52f6f962542
Signed-off-by: Matus Fabian <matfabia@cisco.com>
|
|
multiplexing support:
- set request index as connection_index in application session
- do not allocate app session/http req immediately in
http_ts_accept_callback but wait for request/stream opening
HTTP version specific data in connection and request ctx:
- opaque pointer in http_conn_t
- req_pool manged entirely by specific version engine
Version specific configuration:
- added name to http_engine_vft_t
- added unformat_cfg_callback (optional) to http_engine_vft_t
Type: improvement
Change-Id: Ib43f0489337a222a68b0f81d45cb2e64b2c606c0
Signed-off-by: Matus Fabian <matfabia@cisco.com>
|
|
Type: test
Need to use vcl_test_client and server for deb pkg installation.
After that it is possible to use for the new one CSIT testsuit.
Change-Id: I5a374e47e9c0d0fa3920f26f6dcc9adfbc05103b
Signed-off-by: Ivan Ivanets <iivanets@cisco.com>
|
|
By the DPDK documentation rte_eth_tx_prepare() call is required
to prepare the NIC and to validate the mbufs.
This would fix hardware bad UDP checksum generated by ena driver issue.
As performance will be impacted depending on the driver tx_prepare() callback
I defined a tx-prepare flag that would be enabled by default on ena driver but
could be enabled in the configuration if need for others drivers.
Note: This option would normally be exclusive with intel_phdr_cksum as driver's
tx_prepare would normally cover this usage.
Type: fix
Change-Id: Ic7c21682f7bd92b35bd9b1028129709baa2a64d4
Signed-off-by: Nicolas PLANEL <nplanel@cisco.com>
Signed-off-by: Nicolas PLANEL <nplanel@gmail.com>
|
|
- moved HttpUnitTest to unittests_test.go
- unit tests will only run on coverage builds
- added "SUCCESS" message to some unit tests
Type: test
Change-Id: I35a0510d0ac2899d85f2aa6875a738b280e9b89c
Signed-off-by: Adrian Villin <avillin@cisco.com>
|
|
Add a check so that ICMP type is verified to be an error before parsing
inner payload. If it's not an error, then the inner payload is not there.
Type: fix
Fixes: 46d0ff3945
Change-Id: I5c7d8ddacb347ec030784f349064e66d63cd525e
Signed-off-by: Klement Sekera <klement.sekera@gmail.com>
|
|
Type: fix
Change-Id: Ia1e61a7ed633ad9708bd02fbaf4a39fe1a0ca1f3
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Type: fix
Change-Id: I937d5d8c89bd1e0cc6ff396df6e81ff96755702c
Signed-off-by: Maxim Babichev <maxim@netgate.com>
Signed-off-by: Alexander Skorichenko <askorichenko@netgate.com>
|
|
Format the hardware offload info the same way than other protocol
to avoid confusion when reading logs, like this :
PKT_TX_SCTP_CKSUM (0x0000) SCTP cksum of TX pkt. computed by NIC
PKT_TX_TCP_CKSUM (0x0000) TCP cksum of TX pkt. computed by NIC
PKT_TX_UDP_CKSUM (0x0000) TX UDP cksum
Type: improvement
Change-Id: Icd16a484eabe36af4488352312245d638d7dff4b
Signed-off-by: Nicolas PLANEL <nplanel@cisco.com>
Signed-off-by: Nicolas PLANEL <nplanel@gmail.com>
|
|
enable HTTPS for client apps
Type: improvement
Change-Id: I2ca8b926771a350863cca81729102faf6ee9c874
Signed-off-by: Matus Fabian <matfabia@cisco.com>
|
|
Type: fix
Change-Id: If77469fbdf2b95b0c546a2d1de1cc8663464e1fc
Signed-off-by: Semir Sionek <ssionek@cisco.com>
|
|
Better handling of multi-threaded applications that share sessions.
Type: improvement
Change-Id: Id69bcb1a4b1d67aab020beefdb2fa196ec2ee108
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
One less state in state machine
Type: improvement
Change-Id: Ib6445a425b1e2d5a957318a94c3f132cddd8370b
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
We are currently testing only active open from default to non-default
appns. Add connect test from non-default to default appns and make sure
transport cleans up ports.
Type: improvement
Change-Id: Id0f05aa295ac175b549d8035eb530dbb9a15d85d
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Type: fix
Change-Id: I0b0cc664cec2643a543a39cc7482292604381bf7
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
reset hs->data_len before handling new request
Type: fix
Change-Id: I5b02f8ec418c56e8ec607c5bc5c8716d53e5376f
Signed-off-by: Matus Fabian <matfabia@cisco.com>
|
|
Move all existing host stack scripts under extras/scripts/host-stack
Also add scripts for iperf/vcl performance testing
Type: improvement
Change-Id: Ia79b6bd061db49f61a0e4c6577cf45afe0323eb1
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
In case of sparse message ids due fixed offsets, length of
the message table is less than max message id, causing
"IndexError: list assignment index out of range" exception
in _register_function() due "self.id_msgdef[i] = msg".
Unlike shmem api, socket api needs to use max id.
Type: fix
Signed-off-by: Vladislav Grishenko <themiron@yandex-team.ru>
Change-Id: Ib777db9dabc3a5a3ff83f07ec211cf2fb3c15cf0
Signed-off-by: Ole Troan <otroan@employees.org>
|
|
Type: feature
Change-Id: Id2c6f6b5747c1f676048642e277eb66850f728b7
Signed-off-by: Matus Fabian <matfabia@cisco.com>
|
|
Dump more cpu and reftime fields in clib time struct. Example:
DBGvpp# sh clock verbose
Time now 6.890267
cpu time 6.890267 now 22333609568369233 last 22333609568350337 since start 19981775480
reftime 6.890268 now 1740958292.416887 last 1740958285.526620 init 1740958285.526620
error 0.000000, clocks/sec 2900000000.000000, Sun, 02 Mar 2025 23:31:32 GMT
Time last barrier release 3.291883104
0: Time now 6.890387
cpu time 6.890387 now 22333609568706225 last 22333609568696497 since start 19982121640
reftime 6.890387 now 1740958292.417007 last 1740958285.526620 init 1740958285.526620
error -.000001, clocks/sec 2900000000.000000
Thread 0 offset 0.000000000 error 0.000000000
1: Time now 5.695180
cpu time 5.695180 now 22333609568926421 last 22333609568920073 since start 16516023212
reftime 5.695181 now 1740958292.417084 last 1740958286.721903 init 1740958286.721903
error -.000001, clocks/sec 2900000000.000000
Thread 1 offset 1.195283275 error -.000000114
Type: improvement
Change-Id: I383b9b1a2ef64a731bbc335cd512d565d47f0cd9
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
- It is now possible to use "proto://ip4:port/target" or
"proto://[ip6]:port/target" format.
- Updated http_client and related tests to use the new format
Type: improvement
Change-Id: Ic6afd8c66eddca2ab1d7afc034e193441c34f8ee
Signed-off-by: Adrian Villin <avillin@cisco.com>
|
|
Type: fix
Change-Id: I3e484ed3447dde3540c39b2c5c5ce26329c83340
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
1) header encoding without dynamic table
2) serialization of response header block
Type: feature
Change-Id: I7ec470310e5aec0f8055492e92682261b4af5e81
Signed-off-by: Matus Fabian <matfabia@cisco.com>
|
|
If reassembly is enabled, make sure to apply ABF after reassembly.
Type: fix
Change-Id: Ic12ccba34a9e4d13caba1821e8175ee5adc7f8c3
Signed-off-by: Benoît Ganne <bganne@cisco.com>
|
|
This patch adds support to mirror LACP packets between host and phy.
It is needed for the Sonic-VPP project to support LAG and allow Sonic
to run LACP in the control plane.
The change has 3 parts:
(1) Converted lip_punt_node to lip_punt_xc_inline, which now supports
the creation of two distinct nodes: lip_punt_node and lip_punt_xc_node.
lip_punt_node retains its original punt functionality.
lip_punt_xc_node supports both punt and x-connect between host & phy.
(2) Add 2 new API (and corresponding CLI) to the linux-cp plugin:
(A) lcp_ethertype_enable ("lcp ethertype enable <ethertype>")
(B) lcp_ethertype_get ("show lcp ethertype")
(3) Add UT to test the new functionality and API/CLI for LACP and LLDP.
Type: improvement
Change-Id: Iab66e3b29351dcf2c471babd4f1ef4bdd19da46e
Signed-off-by: Akeel Ali <akeelapi@gmail.com>
|
|
In the file handler, squash groups of forward slashes during path
sanitation to minify the risk of running out of memory.
Type: fix
Change-Id: Ic29d691f876b891ff588157851334162b4e3c5e3
Signed-off-by: Semir Sionek <ssionek@cisco.com>
|
|
1) parsing of binary format (RFC7541 section 6)
2) simple dynamic table implementation
3) parsing of request header block
Type: feature
Change-Id: If43e175a0643f9731c15efc412a82345d9e33cee
Signed-off-by: Matus Fabian <matfabia@cisco.com>
|
|
Check only PING_RESPONSE_IP4 and PING_RESPONSE_IP6
in ping binary API so it doesn't block other binary API
requests while working. (current version of ping binary API
can read other events like SOCKET_READ_EVENT, for example).
Type: fix
Change-Id: Ie7c5c92322af28633680c9c0d60fed739d4e65a0
Signed-off-by: Nikita Skrynnik <nikita.skrynnik@xored.com>
|
|
UNUSED_VALUE warning fixed.
Type: fix
Change-Id: Idd1f0fab96f252f2081d5c268ecc722223289477
Signed-off-by: Piotr Bronowski <piotrx.bronowski@intel.com>
|
|
Move vxlan-gpe folder under vnet to the plugin folder,
update cmake configuration and header paths,
and add plugin.c to register plugin.
JIRA: VPP-2059
Type: improvement
Change-Id: I31b6d326276c4aa684fcdcf8443ef349f7816a6d
Signed-off-by: lajoskatona <katonalala@gmail.com>
Signed-off-by: Nicolas PLANEL <nplanel@cisco.com>
|
|
Type: fix
Change-Id: I77c2d8e64ed463b1c8e84e0fca8d79684a88714c
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Trace both packets generated by dispatching sessions and those that have
been enqueued by other nodes, e.g., syns, fins, acks enqueued by tcp
input and output.
Because not all buffer sources know the session, remove for now the
session index from the trace. Nonetheless, next node on path will print
it, so no information is lost.
Type: improvement
Change-Id: Id69094fbf00e6bc8f98095e90cdf20e2b7f0aeda
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
When key not set to ~0 clib_bihash_key_compare_vec8_8 do not skip empty
slot during search and we crash.
Type: fix
Change-Id: I48f510bceffcfb5e4851dd757cad055a6f7f12d0
Signed-off-by: Matus Fabian <matfabia@cisco.com>
|
|
RFC7541 section 5, Huffman encoding included
Type: feature
Change-Id: I5a024f654610729b8f3f53734aa242bb4b5a25fe
Signed-off-by: Matus Fabian <matfabia@cisco.com>
|
|
IPSec traffic may be sent encapsulated inside UDP packagaes.
In case of esp packgaes decryption is required (according to defined policies),
whereas IKE traffic should be bypassed (relevant policy needs to be defined).
With this patch required behaviour is provided.
Type: feature
Change-Id: If99c7bf121db881c0bdf2b45e6fdca87c0d872a5
Signed-off-by: Piotr Bronowski <piotrx.bronowski@intel.com>
|
|
In case of ipv6 addresses spd did not support bypass and discard policies.
This change introduces missing implementation
in the same way as it was implemented for ipv4.
Type: feature
Change-Id: Idad974655b209d946414d7d85037d0783cde7db3
Signed-off-by: Piotr Bronowski <piotrx.bronowski@intel.com>
|
|
fib_table_get_flow_hash_config accepts fib_index, not fib_table.
Type: fix
Change-Id: I0372ca1b6caab4a34bc0590f9856d89deff6ee90
Signed-off-by: Artem Glazychev <glazychev@mts.ru>
|
|
Type: improvement
In certain use cases, the underlying pools expand by allocating a new,
larger pool and copying the existing elements into it. This process
can be time-consuming, leading to slower control plane configurations,
especially when a large number of elements are already present.
This patch allows users to pre-configure some of these pools through
startup.conf.
It also fixes alignment for ip4 mtrie.
Signed-off-by: Mohsin Kazmi <sykazmi@cisco.com>
Signed-off-by: Benoît Ganne <bganne@cisco.com>
Change-Id: Ib0f1d40e3efb8b4fce989219196c718d6834498a
|
|
Type: improvement
Change-Id: I86497255cd2a73e37ae9be61dcce3a27199c552f
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Separate exception state checks, e.g., no connection or closed, from
segment validation. Segments with no ack, rst, syn flag should not be
received in established node. Still, leave the check in for now.
Type: improvement
Change-Id: I7ceb01d7133f3a571e18721b6e51ff79f533f8cb
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Type: improvement
Change-Id: I9a0437dcfaf5e6930bb7fa057866ea36e7ca328f
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Type: improvement
Change-Id: I03f0b6137db6780f2c2935df90e98acf4bd471f9
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Type: improvement
Change-Id: Idfbaf56e3b56e77c8deaca9d3e41f7a78d8c4e0b
Signed-off-by: Damjan Marion <damarion@cisco.com>
|