Age | Commit message (Collapse) | Author | Files | Lines |
|
Type: test
Change-Id: I299d023ce59b931697352c54e4796ba0af2ab1af
Signed-off-by: Adrian Villin <avillin@cisco.com>
|
|
In Arm platform, targeted prefetching[1] is supported. Adding this in infra
to use it in future optimisation.
[1] https://developer.arm.com/documentation/101458/2404/Optimize/Prefetching-with---builtin-prefetch
Type: improvement
Change-Id: Id3e2a274628289707840a6b2ba4b83c5e4add60b
Signed-off-by: Niyaz Murshed <niyaz.murshed@arm.com>
|
|
Type: fix
Change-Id: If07a6a420d467e6305f166ed250f84befe9f983d
Signed-off-by: Steven Luong <sluong@cisco.com>
|
|
Type: fix
Change-Id: Id8542ad74389eca4cc7903cd455f556743a5e0cf
Signed-off-by: Steven Luong <sluong@cisco.com>
|
|
Allow other apps to own transport sessions for tls.
Type: improvement
Change-Id: Ida04cef1dc49acb7117cbed8657e26890bf43d2a
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Invoking the API session_rule_add_del to add a rule causes VPP to crash
if session is not enable.
Type: fix
Change-Id: If8de83f1064bc60d20461e87889b7c3393ad2af1
Signed-off-by: Steven Luong <sluong@cisco.com>
|
|
lock the vrf table when adding an application namespace and
unlock the vrf table when deleting an application namespace.
Free the session table when no more application namespace
uses it anymore to avoid memory leaks.
Type: fix
Change-Id: I10422c9a3b549bd4403962c925e29dd61a058eb0
Signed-off-by: Steven Luong <sluong@cisco.com>
|
|
Introduce "none", "all" and "failed" options for --decode-pcaps
parameter. Keep "failed" as default to be consistent with current
behaviour. Add missing documentation to test/Makefile and passthrough to
Makefile.
Rationale: running tshark binary takes about 100-150ms and if there are
thousands of pcap files, it takes minutes to decode them. This might not
be desirable if rerunning the tests repeatedly during development.
Type: improvement
Change-Id: Ie033521d51d18b9d499b9bc40fe6eff21c94622d
Signed-off-by: Klement Sekera <klement.sekera@gmail.com>
|
|
An issue is noticed when the link state of the ethernet interface is
toggled to down and again brought up. The link status shows "up" in the
"show interface", but shows "down" in "show hardware-interface".
This issue does not occur when the link is brought up for the first
time, but only manifests when the interface is brought down and up again.
This happens because the device status and VNET_HW_INTERFACE_FLAG_LINK_UP
aren't correctly updated during the port stop, preventing them from
being updated during the port start.
This patch fixes this issue in the port stop function.
Type: fix
Fixes: 01fe7ab88e
Signed-off-by: Alok Mishra <almishra@marvell.com>
Change-Id: Idbc5a6c31409a3d1c5b19e14a3c9509291175677
|
|
Check and skip VPP_EXCLUDED_PLUGINS tests for most of plugins.
Type: improvement
Signed-off-by: Dmitry Valter <d-valter@yandex-team.com>
Change-Id: I23fd3666729251c639aa8da72a676058e3f5bb4e
|
|
clang-format-11 is used by default in checkstyle.sh while it can be
overridden to higher clang-format version. On Ubuntu 24.04, the minimum
version of clang-format is version 14. Set the environment variable
explicitly in Makefile to overwrite the clang-format version.
Before:
$ make checkstyle
extras/scripts/checkstyle.sh: line 41: --version: command not found
make: *** [Makefile:775: checkstyle] Error 127
After:
$ make checkstyle
Ubuntu clang-format version 14.0.6
*******************************************************************
* CHECKSTYLE SUCCESSFULLY COMPLETED
*******************************************************************
Type: fix
Change-Id: Iba3e4739c9376701577b2f37e070d199b195e7d0
Signed-off-by: Jay Wang <jay.wang2@arm.com>
|
|
CMAKE_C_COMPILER_TARGET doesn't need to be set when it's not cross
compiling, and based on wrong identation of that line I suspect it was
intended to be inside CMAKE_CROSSCOMPILING conditional.
With this line moved inside, it ends up being the only place where
COMPILER_SUFFIX is used, so, move its declaration there as well.
It will not bring any change to currently supported systems but will
make it easier to build VPP on systems running musl libc.
Type: fix
Change-Id: Ia8e16d9e8ad9c987ddf246dd0e3f8636ccad1362
Signed-off-by: Renato Botelho do Couto <renato@netgate.com>
|
|
- support for ubuntu-24.04 CI jobs
Type: make
Change-Id: I225d72e22207283de9979cce75d8d208ab99feb5
Signed-off-by: Dave Wallace <dwallacelf@gmail.com>
|
|
When a FIB entry has multiple sources installed by covering inheritance,
the first walk installs the covering source, which can be worse than the
already used one (hence the source is not activated). If a covering path
is added and a new fib cover walk is done, the deactivated source is
compared to itself. This operation calls for a reactivation of the source which was not activated in the first place.
Type: fix
Change-Id: I2639401bed14eb5a2cb003664c04d6c1f9130de9
Signed-off-by: Maxime Peim <mpeim@cisco.com>
|
|
Avoid to do fixup either if TTL or MTU check fails. It rewrites the
icmp header copied in case the ip6-icmp-error ran before.
Type: fix
Change-Id: If043d1cae35c41ebf70ba979883119604bee883d
Signed-off-by: Maxime Peim <mpeim@cisco.com>
|
|
VPP IPv6 FIB relies on 2 bihash for all its VRFs: 1 for control plane
and one for data plane. Because of that, control plane VRF walks
time grows with the number of VRFs as it has to go through all the VRFs.
In order to improve control plane performance with large number of VRFs,
the control plane data structure is changed from a global bihash to a
per-VRF hashtable, similarly to IPv4. Contrary to IPv4, there is only 1
hashtable per VRF and not 1 hashtable per prefix-length per VRF, which
is a compromise between memory consumption and single VRF scaling.
Type: improvement
Change-Id: Ib738a44adb88e46c95c4c62c853ee1f1392c7187
Signed-off-by: Benoît Ganne <bganne@cisco.com>
|
|
In some cases we do not need multicast support. Making it optional helps
scaling to high number of VRFs, by reducing the control plane operations
and memory consumption.
Type: improvement
Change-Id: Ib34ed3fe2806e2f4624981da4e4a3c49c69f70be
Signed-off-by: Benoît Ganne <bganne@cisco.com>
|
|
Destroying a non-empty fib table can lead to difficult to diagnose bugs,
like dandling fib entries preventing entries to be inserting to the
mtrie.
Always log an error to help diagnose those issues.
Type: improvement
Change-Id: I2c771a80595035b440931b74cca4429af22161a2
Signed-off-by: Benoît Ganne <bganne@cisco.com>
|
|
Type: test
Change-Id: Iaed08fe23a63562c99012bd469ca3f3271be2564
Signed-off-by: Adrian Villin <avillin@cisco.com>
|
|
We call application_namespace_cleanup to remove the application which
has a reference to the application namespace when deleting an application
namespace. When we find the matching application namespace, we queue
the ns_index instead of app_index to look for the application.
This may lead to a crash.
Type: fix
Change-Id: I89c7f8d1681b4c7c1bcb16641217230a26dbe582
Signed-off-by: Steven Luong <sluong@cisco.com>
|
|
Type: test
Added suite to verify that VPP launches with provided
CPU pinning configurations. CPU configuration is
specified per-test.
Change-Id: Ic283339676d3b24636fc21156a09a192c1a8d8da
Signed-off-by: Hadi Rayan Al-Sandid <halsandi@cisco.com>
|
|
Running VPP on a NUMA system with 5.19+ kernel outputs the following
error messages. 'show physmem' command confirms that VPP falls back to
using normal 4K pages instead of the preallocated 1G huge pages. The
root cause is that VPP uses move_pages()[1] to get the huge page node
information. However, this misbehaves on the 5.19+ kernel due to changes
introduced in its implementation[2].
Our proposed fix is retry obtaining NUMA node info with
get_mempolicy()[3] only if we see -ENOENT returned in status from
move_pages() and huge pages are used. Additionally, we use mincore()[4]
to check if pages are allocated and in memory to avoid the possibility
of get_mempolicy() falsely allocating a new page.
buffer [warn ]: numa[1] falling back to non-hugepage backed buffer pool ()
vpp# show physmem
used-pages 2 reserved-pages 16 default-page-size 1G lookup-page-size 4K
arena 'buffers-numa-0' pages 1 subpage-size 1G numa-node 0 shared fd 5
arena 'buffers-numa-1' pages 1 subpage-size 4K numa-node 1 shared fd 6
[1] https://man7.org/linux/man-pages/man2/move_pages.2.html
[2] https://lore.kernel.org/linux-mm/91da2c3b-96f1-bb03-8fff-4c38f31cb9be@huawei.com/
[3] https://man7.org/linux/man-pages/man2/get_mempolicy.2.html
[4] https://man7.org/linux/man-pages/man2/mincore.2.html
Type: fix
Signed-off-by: Jay Wang <jay.wang2@arm.com>
Change-Id: Ia423745423bb080404292333ef95455a4950ce0a
|
|
Type: refactor
Change-Id: I3d7ff64e2e54b113f8d4b26f2dbf5b9c67b80976
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Type: improvement
Change-Id: I5a4a33604ecb8a29ef6932cc5826835d3bec7f23
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Type: test
Change-Id: I2730debc7ea9256bf27208a2662f920edcd6826c
Signed-off-by: Dave Wallace <dwallacelf@gmail.com>
|
|
- Release build runs on numa node0, debug on node1.
Using the last digit of a build number to reserve 4 cores per test
mmeans we can run 20 jobs (10 release, 10 debug) on the same machine,
assuming we have 111 cores available (not counting core 0).
Can be increased if needed, there are still some cores left.
- Added separate numa aware cpu allocation
- Added CPU0=true|false (useful for users with 4c/8t)
Type: test
Change-Id: Iba8e492a4e01a7f457e49112303887a2a27f6af9
Signed-off-by: Adrian Villin <avillin@cisco.com>
|
|
Type: improvement
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: I3c79d16f6a19767d990e8a4683c296219b559ccd
|
|
Type: fix
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: Iadc6ec58f5fc7fbff62e4fd31d509557b60395cb
|
|
This change is used to add cli command for macip acl rules.
Cli commands added for addition and deletion of rule.
Type: improvement
Change-Id: I28adba0193a904f6c130c81cd19bb6de47f517a9
Signed-off-by: Sanjyot Vaidya <Sanjyot.Vaidya@arm.com>
|
|
Also rework the logic so the skipping of marked Ubuntu 22.04 occurs at framework level
Leave debian11 special cases as-is.
Type: fix
Change-Id: I481eb32cd1a0860935482e9f930ced409da653c9
Signed-off-by: Andrew Yourtchenko <ayourtch@gmail.com>
|
|
If session is not enable, type in the CLI session disable cause VPP to crash.
Type: fix
Change-Id: Ib55b7430b7ce521a40025d58fa18b765f5129fd5
Signed-off-by: Steven Luong <sluong@cisco.com>
|
|
check return value after strstr()
Type: fix
Change-Id: Ib6eb82b6a52c7e08785f6deb73e6cc7a10e7f61b
Signed-off-by: Steven Luong <sluong@cisco.com>
|
|
When toggling session disable and enable, memory leaks found on various
transports.
Don't create the timer wheels on enable when they are already creaded.
Don't create the spin locks when they are already created.
Add session enable disable test in session unit test.
Type: fix
Change-Id: I08988a96bbf2d67a67c743b50b63f5b1e27402a3
Signed-off-by: Steven Luong <sluong@cisco.com>
|
|
- as per Florin's request
Type: test
Change-Id: Id76fda0e2dfac7e58b703a8d6f096aa7b5be31c7
Signed-off-by: Adrian Villin <avillin@cisco.com>
|
|
Type: docs
Change-Id: Id47dbab5e5661c76e6ff41237bbca5b5b15506c4
Signed-off-by: Andrew Yourtchenko <ayourtch@gmail.com>
(cherry picked from commit 6e8b350a0154caee6e9f037ae4938b83fb5dcf20)
|
|
Use standard macros CPU_ISSET_S and CPU_ZERO_S and make code more
portable.
Type: improvement
Change-Id: I8e6a0fd79b90de8c6c33cad8882680eae2ca50ef
Signed-off-by: Renato Botelho do Couto <renato@netgate.com>
|
|
When custom netns is used we don't need to populate the mapping
ifindex->sw_if_index otherwise netlink events in default-ns can apply settings
to wrong interfaces. Most of the lcp functions however wasn't working for such
interfaces and neither it was fixed by this patch.
Type: fix
Change-Id: I74a8a4e332753f9a40fc291d489dfd7fc51cbef3
Signed-off-by: Stanislav Zaikin <stanislav.zaikin@46labs.com>
|
|
Type: refactor
Change-Id: If3a195f1ea600bfd621717f333af1b1273ab0807
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
When the rule table is allocated, the root node is allocated. But it
cannot be deleted by the add_del API, or the table is useless.
When the table is free, the root node is not free and there is memory
leak. Let's add pool_flush when the rule table is free.
Type: fix
Change-Id: I58c3e040cd101c7db501d99a373ad78d85321b8f
Signed-off-by: Steven Luong <sluong@cisco.com>
|
|
RFC6935 section 5 states that, by default, the UDP checksum must be
computed when originating an IPv6 UDP packet. The default behavior
may be overridden when conditions defined by RFC6936 are satisfied.
However this implementation does not satisfy all the conditions so
the checksum must be computed.
After ESP encryption the packet is an IPv6 UDP packet so set the
l3_hdr_offset and l4_hdr_offset values, and set the UDP_CKSUM
offload flag in the buffer.
Type: improvement
Co-authored-by: Cian Ferriter <cian.ferriter@intel.com>
Change-Id: I9f8c434c9fe9dbddd8890d5ae366984bfcf34067
Signed-off-by: Jeff Shaw <jeffrey.b.shaw@intel.com>
|
|
Type: fix
Fix case where no error code would be returned if vpp was launched,
but then exited due to invalid configuration.
Change-Id: I54d526629a2ee0206227615ffb6cb658779f93a0
Signed-off-by: Hadi Rayan Al-Sandid <halsandi@cisco.com>
|
|
Adding prefetching to openssl plugin improves both sync and async mode
ipsec performance by more than 5% on N1 ampere.
Sync mode (1420b)
core count old (MPPS) new (MPPS) %diff
1c 0.972 1.01 3.90
2c 1.91 2.02 5.87
3c 2.86 3.04 6.03
Async mode (1420b)
core count old (MPPS) new (MPPS) %diff
1c 1.296 1.37 5.70
2c 2.58 2.753 6.70
3c 3.74 3.9 4.27
6c 7.52 7.832 4.14
Type: improvement
Change-Id: Ieef22c37e1330ac9f8b7e09a25c24162516b6c26
Signed-off-by: Niyaz Murshed <niyaz.murshed@arm.com>
|
|
Type: test
Change-Id: Ie525636c6299a8306cba45e72f8ee6c9da6d6e4f
Signed-off-by: Matus Fabian <matfabia@cisco.com>
|
|
Type: test
Change-Id: I327fa1a4ea23a3af3aa33e5260367426a11e7b4f
Signed-off-by: Adrian Villin <avillin@cisco.com>
|
|
Older distributions like Ubuntu 20.04 uses an old version
of cmake that does not handle absolute paths.
Leave .api dependency handling like before, i.e.
not handle interdependency between .api files for those older
builds.
Type: fix
Change-Id: I3f15a4506f6c95fca7a5b00893dc354cbed42a29
Signed-off-by: Ole Troan <otroan@employees.org>
|
|
Since main thread is the only one thread that can modify address
hash, avoid barriers while IP_ADDRESS_DUMP api calls.
Type: improvement
Signed-off-by: Vladislav Grishenko <themiron@yandex-team.ru>
Change-Id: I9c35400bdef9fb08e81d859c8f2e3bdaa5468f4b
|
|
- when running in the CI, cache the docker build layers
in the local filesystem to allow docker executor images
to contain cached docker build layers
Type: test
Change-Id: Ie728a8370d3fb8144d01dff566aaa846ca6fd81b
Signed-off-by: Dave Wallace <dwallacelf@gmail.com>
|
|
Type: fix
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: Ica78044b044a78714dd9947620250f13c63320fb
|
|
Type: fix
Change-Id: I3a5c04b14bf1156376e2acece69f8256d00f261a
Signed-off-by: Maxime Peim <mpeim@cisco.com>
|
|
Fixes: dd6fb60f1794fc08ec40598a67dc70f942c200d1
Type: fix
Change-Id: I2429715a954361ceea969191493c15bef21e2040
Signed-off-by: Ole Troan <otroan@employees.org>
|