Age | Commit message (Collapse) | Author | Files | Lines |
|
Signed-off-by: Peter Mikus <peter.mikus@protonmail.ch>
Change-Id: Id3f0a4a26404951bc7ce82655538b72848508e18
(cherry picked from commit f5275f845fed64aaf8f86913e289b31b410837f2)
|
|
Signed-off-by: Peter Mikus <peter.mikus@protonmail.ch>
Change-Id: I91a3136a29192fb25ec4845e0924b96977f591f5
(cherry picked from commit a410c9fab12fd5d2087d5429860dfd9bcc49986b)
|
|
Change-Id: If5003d4fb718dc1bf76cb6e049b00783e03de8bb
Signed-off-by: Tibor Frank <tifrank@cisco.com>
|
|
Change-Id: I5b983aa249dd5ac62f78e6f980ab81f270234b0f
Signed-off-by: Tibor Frank <tifrank@cisco.com>
|
|
Signed-off-by: Peter Mikus <peter.mikus@protonmail.ch>
Change-Id: I28f2752f725c9816a557667033bf7d656fb0d414
|
|
Signed-off-by: pmikus <peter.mikus@protonmail.ch>
Change-Id: I95171ab59dfe1b5a98c36a839233d14e9ddb7ce5
|
|
Change-Id: Id2535d02d103db58888dac7733d5aded1e7c5d78
Signed-off-by: Tibor Frank <tifrank@cisco.com>
|
|
+ List added and edited perf tests.
+ List new device NICs and tests on them.
+ Copy known device issues, even thougth new jobspecs avoid them.
Change-Id: I3e791c6d6395a4d9ac720fb0c9291464cf6d7395
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
Fixes: d29977e35bb15cd7df21c082432f03eab87e2e66
Change-Id: I1a20ded1181f5a3b33a0f6640554a939218d9791
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
Change-Id: I53d24a291907c862e2214e9152cede8f0ac05345
Signed-off-by: Tibor Frank <tifrank@cisco.com>
|
|
Change-Id: I038639ad010935d6144c79acb25e0e357beb3e96
Signed-off-by: Tibor Frank <tifrank@cisco.com>
|
|
The -ar- suites had everything edited,
except that one agrument that would actually enable anti-replay.
Fixes: b9dff484de0725bca3f65739519d952e3be55cbe
Change-Id: Ib2c590a9be5271348ee15d4e4ab1b7c663559477
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
Change-Id: Ic7c09ffab66e7ccd604653ea26c35ceaffa65094
Signed-off-by: Tibor Frank <tifrank@cisco.com>
|
|
+ Only for dpdk-plugin and primary NIC.
+ Prefer Intel, as that seems to have more stable performance.
+ Only one testcase per suite: 1518B 1c.
+ Add full set of 10k AES256GCM tests where not present already.
+ Remove multi-core ipsec AVF tests from 3n-icx to keep duration small.
- Performance of those was limited by CSIT-1948 anyway.
Change-Id: I9166f15b44071dafe89ac8ca7982f659006027ba
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
+ This mainly uncomments a bucnh of 3n-alt xl710 ipsec tests.
+ Otherwise the difference was quite small.
Change-Id: I8b67fcce157f36ed79e9692a0581e1115b3f6886
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
+ Add suite with UDP encap.
+ Add suite with anti replay enabled.
+ Add new enums where needed by the new suites.
+ Apply trimming in enum_util to support "3DES".
+ All 10ktnl suites written and tested.
+ New robot tags added.
+ Fix one comment from the parent.
Change-Id: I2581814dbb327891d8658dd009c4e52ffd318e3b
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
+ Add l2xcbase.
+ Add 200k scale, both ip4 and ip6.
+ Test 200k and l2bdmaclearn also on:
+ Primary NIC with secondary driver.
+ Secondary NIC and its primary driver.
+ Add 1518B memif test cases, for 2n-spr also with DMA (including 64B).
+ This needed an edit in autogen.
+ Switch crypto tests to drv_vfio_pci as AVF has bad RSS on those.
+ Add two SRv6 cases to 3n-icx as it does not support hwasync crypto.
+ Repair name of UDP ASTF tests.
Change-Id: I91e20ec80dc4984c55d39b5f5ad38f26b48da33e
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
Avoiding any predictable memory access patterns
is more important than similarity between 2p and 6p profiles.
+ Seed values now match parts of IP source address.
Change-Id: Ib1830cf9d3e7b329e54de19a92ecea77c8baeeb5
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
As hoststack is never tested in container/VM (yet),
configuration there does not check the new flag.
Change-Id: Icac4837efcf035cad0f1e126e1bf90811aba16b2
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
Previously, the flag was sometimes called "jumbo",
sometimes "jumbo_frames". As we never use "jumbo" for things
other than frame size, unify the flag to "jumbo".
Change-Id: Ib688389799a4550ac9f51375f4defc06a02e6f03
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
This reverts commit 4e897a7a91e8b4782da05a4c906392e62c8f8ab5.
Reason for revert: Container/Qemu is not conditioned.
Change-Id: I497caa2db24803f40811b3c60c359b4772cacda7
Signed-off-by: Peter Mikus <peter.mikus@protonmail.ch>
|
|
Change-Id: I169b06433ae962811aafa91ce90fe169b0bc4c71
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
Change-Id: I1bb9b3af12737203d45f981b80be88290d2c7be4
Signed-off-by: Tibor Frank <tifrank@cisco.com>
|
|
Move DRV_VFIO_PCI after TXQ_SIZE_0.
Upcoming suites need more space in the preceding line.
131 suite files edited, luckily doable by two commands:
$ sed -i 's/ | DRV_VFIO_PCI//' *.robot
$ sed -i 's/ | TXQ_SIZE_0/ | TXQ_SIZE_0 | DRV_VFIO_PCI/' *.robot
Change-Id: Idf30d20d321922718d523a216056d22ae67e50aa
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
Signed-off-by: pmikus <peter.mikus@protonmail.ch>
Change-Id: I3ef110e8fec2fa8abafead1392a1b2af5e59f803
|
|
Signed-off-by: Peter Mikus <peter.mikus@protonmail.ch>
Change-Id: I1907fcb88dff0bb62cc6f5a8bf3ea58e3ad1578d
|
|
Signed-off-by: Peter Mikus <peter.mikus@protonmail.ch>
Change-Id: I6951c4fc9f96a0acf44ba3201c91c9cb903889d7
|
|
This makes most scale hoststack tests correctly use both sibling cores.
- Except quic multi-client ones, which start failing instead.
- To be fixed later.
- VPP startup.conf looks correct to me.
- All quic tests suffer from CSIT-1935 so no big loss.
Change-Id: If22c01651140865e188733039890f1fb0ceaa57a
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
While adding runtime actions to hoststack tests is not easy,
adding pre and post actions for stats is easy.
The main hoststack measurement is treated as the stats trial,
the same way as the 10 measurements in MRR tests are.
+ Remove the previous partial info via "show" CLI commands.
- As in other VPP tests, stat telemetry is not exported.
Change-Id: Ib567aa810418568520d4d543c6d50431e85a3269
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
Change-Id: Ie3a61153c3e286f196617be438c2056a001b6faa
Signed-off-by: Tibor Frank <tifrank@cisco.com>
|
|
Signed-off-by: Peter Mikus <peter.mikus@protonmail.ch>
Change-Id: Ibd3d0a1cdce67d17bfbe6fadfc1311af420a6c50
|
|
This reverts commit 1776b3a1a483fd87170dfc0bf2eeb7eb610783bc.
Reason for revert: .
Change-Id: Ibad4684d16c4a348f27371cb2c4108f07cdaed21
Signed-off-by: Peter Mikus <peter.mikus@protonmail.ch>
|
|
Change-Id: Iba15dc18b9e49dee0776e2897e45264a20ff2817
Signed-off-by: Tibor Frank <tifrank@cisco.com>
|
|
Change-Id: Id597a02e3d8586f4c712ebf9eaddc46315b2a844
Signed-off-by: Tibor Frank <tifrank@cisco.com>
|
|
Change-Id: If6f1c306c6c5d72708e447825f17debedbaa4b4f
Signed-off-by: Tibor Frank <tifrank@cisco.com>
|
|
Signed-off-by: Peter Mikus <peter.mikus@protonmail.ch>
Change-Id: I5cec80ece019f660bc449349ce78ec2f35c5bd22
|
|
Signed-off-by: Peter Mikus <peter.mikus@protonmail.ch>
Change-Id: Ibd49015bc810a040d28509efc2e4bca26875229e
|
|
Signed-off-by: pmikus <peter.mikus@protonmail.ch>
Change-Id: I4ce20267b4747bf1901b6175e0ec5936b583a510
|
|
Change-Id: Ia4355295e783f29c87db6ed009370bd794d9b4ce
Signed-off-by: Tibor Frank <tifrank@cisco.com>
|
|
Change-Id: I89f2d314ff5ed33a86407489733544f197ad3d87
Signed-off-by: Tibor Frank <tifrank@cisco.com>
|
|
+ add switching on/off the apps
Change-Id: I46e0cd2f214fd6a05ec0b87a433708bc4b6e2eb1
Signed-off-by: Tibor Frank <tifrank@cisco.com>
|
|
Signed-off-by: pmikus <peter.mikus@protonmail.ch>
Change-Id: I601a54e543114957592548618df2f2f4bf790f82
|
|
Some tests were already doing that via Set Test Variable,
this moves the flag into Variables table for every suite.
Change-Id: I54e86f3fd10354dc87c2f04c70bbab771f89f3d2
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
The old limit of 1522 was introduced long time ago.
First, it appeared here [0], where it is correct
as that suite has zero overhead.
The first suite with wrong logic seems to be here [1]
(no "Add No Multi Seg to all DUTs" in 1518B testcase).
And when I was moving that logic to a keyword in [2],
I did not realize it is wrong with overhead.
This Change uses 1800 as the new threshold,
matching the value used for non-jumbo MTU.
[0] https://gerrit.fd.io/r/c/csit/+/2652/12/tests/perf/Bridge_Domain_Intel-X520-DA2.robot#70
[1] https://gerrit.fd.io/r/c/csit/+/4454/96/tests/perf/40ge2p1xl710-ethip4ipsecscaleip4-ip4base-interfaces-aes-gcm-ndrpdrdisc.robot#229
[2] https://gerrit.fd.io/r/c/csit/+/13411/35/resources/libraries/robot/performance/performance_utils.robot#84
Change-Id: Iff3703fcff0e4bbb1a6b10be359fa5ef67fd5422
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
Change-Id: I818cc10e971c85cd8e712c05536b5813bf18749a
Signed-off-by: Tibor Frank <tifrank@cisco.com>
|
|
Signed-off-by: Peter Mikus <peter.mikus@protonmail.ch>
Change-Id: I5ca0fc5832c249553c79dd9c5d4148ba64e61654
|
|
Signed-off-by: Peter Mikus <peter.mikus@protonmail.ch>
Change-Id: Icbcd373ef815fa082d9be022fc201c96df140536
|
|
Some perf suites use a single unidirectional flow of traffic.
On SMT-enabled machines, 1C tests create 2 workers by default.
But if there is only one stream/flow,
only one worker does useful work.
The other worker on the sibling logical core
does busy-wait, which provably lowers the peformance.
Some suites already set smt_used (as a suite variable)
to False for these purposes.
This change edits a lower-level L1 keyword to honor this value,
and adds the suite variable to single stream (and client) hoststack
suites, where the detrimental effect of busy siblings was observed.
Change-Id: I203f1a700d8e6cd84640b9c0b40487cb832908aa
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
+ State frame size as 1518 octets, even if some packets are smaller.
+ This implicitly makes CSIT keywords to apply n-milti-seg.
- Real frame size is currently not important.
Change-Id: I754eac7e7ebf0378efb27149862004e35c387505
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
Change-Id: I8ab964a6196dc3e81a4b7a9e7d8a7f0c0e2a0002
Signed-off-by: Tibor Frank <tifrank@cisco.com>
|