Age | Commit message (Collapse) | Author | Files | Lines |
|
worker-handoff node is universal node which taakes packets
from the input node and hands them over to worker threads.
Currently it supports flow hashing based on ipv4, ipv6 and
mpls headers.
New cli:
set interface handoff <intrerface-name> workers <list>
e.g.
set interface handoff TenGigabitEthernet2/0/0 workers 3-6,9-10
Change-Id: Iaf0df83e69bb0e84969865e0e1cdb000b0864cf5
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Change-Id: I3379517535a98a7a5fbd3173503f288a81378821
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Previous patch was printing wrong data since rewrite node traces
an ethernet packet while other ip nodes trace an IP packet.
This patch introduces different tracing functions for:
- lookup node (where the fib index is valid)
- rewrite node (where an ethernet frame is available)
- other ip nodes (only the ip header is traced)
Change-Id: I5971a2e89ae8668f5aed4a410565a5f27e01fc22
Signed-off-by: Pierre Pfister <ppfister@cisco.com>
(cherry picked from commit 584b99a0120778108bd019b697639fbf3c1505d1)
|
|
Otherwise, if tapcli-rx manages to make it into polled-mode, it will
never leave polled mode. This has been wrong since day 1, sometime in
early 2013.
Change-Id: I124e01a48db5abbc4eb5240c8d59f76dc562e9f2
Signed-off-by: Dave Barach <dave@barachs.net>
|
|
Change-Id: I3f5113533b365366cc3f6ef87858e19879f703cf
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Verbose trace can be enabled with:
trace add dpdk-input 1 verbose
Change-Id: If58cfc561143d5a44f7d11e61d54b6835692a0e3
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
The absence of trace in ip lookup nodes is misleading to many people.
This patch adds ip lookup tracing and therefore contribute to
worldwide happiness.
In addition, this patch makes sure sw_if_index[VLIB_TX] is
considered when tracing the fib_index value. In ip4/6-rewrite,
the value corresponds to the tx interface index. The formatting
function is therefore modified to take that case into account.
Change-Id: I5915f0446a15c45e391eedfdfcedd9057aa6a237
Signed-off-by: Pierre Pfister <ppfister@cisco.com>
|
|
Change-Id: I1325b60b6deadcb51631e178011a31ee70c06cc7
Signed-off-by: Dave Barach <dave@barachs.net>
|
|
Change-Id: I4d7d4a56aba010aa868b1f4c2c4e8db0b4c21fd7
Signed-off-by: John Lo <loj@cisco.com>
|
|
The old ACL CLIs still show up in CLI help which can confuse users.
Change-Id: I9a3722d3d649c4370df6a09b2c07628e7e4aa0f4
Signed-off-by: John Lo <loj@cisco.com>
|
|
Change-Id: I8292628ad359a19e02ff79568d773ad2b1af51ec
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
This commit removes assumption that all packets have
data offset equal to RTE_PKTMBUF_HEADROOM. Some drivers
like fm10K receive packets with different data offset.
Change-Id: I0aba6296458dab9df6fff639b4b827b2084ddc3e
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Remove usage of dummy encap string for output from BD to a tunnel which
has been deleted. Instead, use a node l2-outpt-del-tunnel so that if
there are stale entries in the L2FIB for any deleted tunnel sw_if_index,
l2-output will send packets using this entry to the l2-output-tunnel-del
node which just setup the proper drop reason before sending packets
to the error-drop node to drop the packet.
Change-Id: I590982ee25e924ab74e2855c55c58baf29a9fad4
Signed-off-by: John Lo <loj@cisco.com>
|
|
Locks are used while populating DB, which is once
once per entry. Therefore, it is not in the performance
critical path. Each thread is a PMD, therefore, spin locks
are used instead of mutexes.
Change-Id: I4bc297f73a8f3eafebed1f00e51ec75ca24163f6
Signed-off-by: Shesha Sreenivasamurthy <shesha@cisco.com>
|
|
Change-Id: Iaf72e7ffb2d30799733f641d66bbc74f40a4d6e1
Signed-off-by: Hongjun Ni <hongjun.ni@intel.com>
|
|
experimental
IOAM code. Support dynamically adding options. By default
only process HBH if explicitly configured to. Otherwise we'll just set
ourselves up to be a victim of DOS.
Change-Id: I41cdfdc00aeaa0cf568e4463440b89be761b6b7d
Signed-off-by: Ole Troan <ot@cisco.com>
|
|
This patch introduces following 2 startup options:
num-rx-desc
num-tx-desc
Which can be specified under the dpdk { dev PCI_ADDR {...} } or
dpdk { dev default {...} } .
"show hardware" output is extended to display what is set:
TenGigabitEthernet2/0/0 5 down TenGigabitEthernet2/0/0
Ethernet address 90:e2:ba:96:d0:54
Intel 82599
carrier up full duplex speed 10000 mtu 9216
rx queues 2, rx desc 512, tx queues 2, tx desc 1024
cpu socket 0
Change-Id: Ia832885c8e5cf0eb0575367e97581e7065934753
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
As BVI receive a packet with unicast DMAC from the BD, including unknown
unicast flood packet, the packet should not be L3 forwarded unless its
DMAC matches the MAC of the BVI.
Change-Id: I46e18629c901062592c8ebe3a238c5cfdc1096b4
Signed-off-by: John Lo <loj@cisco.com>
|
|
This change adds support for providing per-interface
parameters in the startup config.
Sample configuration:
dpdk {
dev default {
num-rx-queues 3
num-tx-queues 3
}
dev 0000:02:00.0 {
num-rx-queues 2
num-tx-queues 2
}
dev 0000:02:00.1
}
Change-Id: Ia7d9ae2ac9c4fd9baaa480d061a395f8a421a722
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Sleep <nn> milliseconds after each dpdk input device poll, useful when
oversubscribing CPUs.
Change-Id: I90ad1f21dae7eeeda56bfe845911118aa46f83ec
Signed-off-by: Dave Barach <dave@barachs.net>
|
|
Fix previous change that does not work on little endian machines.
Use inline call ip4_is_fragment() which is endian neutral.
Change-Id: I5a35d89d936650ab6c628dfc388b8c992a74a589
Signed-off-by: John Lo <loj@cisco.com>
|
|
Some small fixes to the srv6 code to bring it toward
conformance with draft-previdi-6man-segment-routing-header-05.
- The first segment needs to remain in the segment list.
- The segment list template needs a space for the ultimate
destination.
- The ultimate destination needs to be inserted into that
space when adding the SRH.
Change-Id: I66db6912e0128da084f14ceca20918ef67ccff79
Signed-off-by: Chris Luke <chrisy@flirble.org>
|
|
The add_del function was not properly checking if a tunnel already
existed; instead it was checking if the given tunnel name existed.
If no tunnel name was given it flat out refused to add a tunnel
even though that is optional.
Cleanup the add/del parameter validation to "do what I expect" it
to do:
When adding a tunnel:
- If a "name" is given, it must not exist.
- The "key" is always checked, and must not exist.
When deleting a tunnel:
- If the "name" is given, and it exists, then use it.
- If the "name" is not given, use the "key".
- If the "name" and the "key" are given, then both must point to the
same thing.
Change-Id: I9b48ae0203f9664cf8af0f7dc49bf480ddec10d5
Signed-off-by: Chris Luke <chrisy@flirble.org>
|
|
Before VPP is enhenced to do IP reassembly of local packets, just set
the protocol path to "experimental" for now so they will be error-punted
as unknow IP protocol.
Change-Id: I2ffefb0b4205357653ba24d80c722cafd5972fba
Signed-off-by: John Lo <loj@cisco.com>
|
|
When creating VXLAN over IPv6 the code was using storage for the
hash key that could later be moved. Since the key is larger than
the word size this was being referenced as a pointer; when the
storage moves that breaks the hash.
Instead allocate dedicated storage for the key.
This patch also includes other minor cleanups, including using
clib_memcpy in places it should be used and some whitespace
fixes.
Change-Id: I579f2cb515853ef56dedcca350fcad08aa6111a9
Signed-off-by: Chris Luke <chrisy@flirble.org>
|
|
Change-Id: Ida11bddb52268e0e8513b7b379eeed6103bd48f1
Signed-off-by: John Lo <loj@cisco.com>
|
|
The loopback interface should use vnet_interface_output_no_flatten so
follow-on buffers of a jumbo packet do not get put on the output frame
and be sent back to ethernet-input or l2-input node.
The replication_recycle_callback() function should not assume follow-on
buffers of a jumbo packet are on the buffer free list.
Change-Id: Ide646a6d9b43e82782c0581ea3022a9e70f82582
Signed-off-by: John Lo <loj@cisco.com>
|
|
- consistent use of hash_*_mem methods.
- tunnel setup copies the key from the wrong location
- remove extraneous storage for keys copied from the vxlan
node that is not applicable here
Change-Id: I419fd30a52ba387104abe467c296de85233823d1
Signed-off-by: Florin Coras <fcoras@cisco.com>
Signed-off-by: Chris Luke <chrisy@flirble.org>
|
|
Change-Id: I5aef12d3a9c8daefff52e5f958c504f5d2ff9fd0
Signed-off-by: Filip Tehlar <ftehlar@cisco.com>
|
|
Check duplication locator and clean after locator when remove it.
Refactoring locator_set code.
Change-Id: Ib83cbcddc7a363a60fa5b6a366203d0dc0ea7ca6
Signed-off-by: Andrej Kozemcak <akozemca@cisco.com>
|
|
Change-Id: Ia6eaaa9a741e84cbd0ff957cfd9a7143a3d0a977
Signed-off-by: Filip Tehlar <ftehlar@cisco.com>
|
|
Change-Id: I4913fe6c4b1280939147887896aea9b79a9f7f10
Signed-off-by: Hongjun Ni <hongjun.ni@intel.com>
|
|
When af_packet signals the kernel that there are packets in the tx
ring with sendto() the kernel sometimes responds with EAGAIN.
Previously the af_packet driver would treat any error from sendto()
as fatal.
Whilst there's not much we can do about this, count the errors
and let's try to not die on the spot or sit in a loop forever.
Change-Id: Id76ba5e07b744f1ed6f348ec838a1ac506a381c9
Signed-off-by: Chris Luke <chrisy@flirble.org>
|
|
Change-Id: I0e985b079da3224f4886e3ee2cece4d046e291eb
Signed-off-by: John Lo <loj@cisco.com>
|
|
Change-Id: I3ed6269f9a307b50b56ce04defc2576a139c6add
Signed-off-by: Filip Tehlar <ftehlar@cisco.com>
|
|
This is preparation work for moving DPDK
EAL and mempool initialization to vlib.
Change-Id: I2dc35aa53edec0e07fc2650d630aa625831154c3
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Change-Id: Ib95338d7055dea9036eddb7dff5ca7ccbcb35639
Signed-off-by: Andrej Kozemcak <akozemca@cisco.com>
|
|
PatchSet4: consolidate code as per comments.
PatchSet3: simplify the code using ip_udp_encap_one/two
PatchSet2: consolidate comments and indent style
Change-Id: Ia8b43f854a46d77e838e198566200ad28fd72472
Signed-off-by: Hongjun Ni <hongjun.ni@intel.com>
|
|
Change-Id: Ib283c75505b787f715e34f5b325dda8a6c2fa125
Signed-off-by: Hongjun Ni <hongjun.ni@intel.com>
|
|
Available only in vpp_lite platform
Change-Id: I09d112af5f7f4521ec25196ecdd8c02c20eedd5f
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
The current location in the vnet dynamic library does not override the dpdk
weak function as intented. Moving this function to the main app allows the
linker to find the non-weak symbol and override it.
Change-Id: I96c6dc5af2e511cc1f84bd66419c5a4ac96a9541
Signed-off-by: Todd Foggoa <tfoggoa@cisco.com>
|
|
When the CLI is used to create an interface, and whend the operation
succeeds, the newly created interface name is printed-out.
The patch includes the following interfaces types:
- AF_PACKET
- Vhost User
- Netmap
- GRE
- L2TP
- MPLS-GRE
- Loopback
Change-Id: Id518c139ec63a261eae81d2ed95c4cd1f10b5157
Signed-off-by: Pierre Pfister <ppfister@cisco.com>
|
|
Change-Id: I8a3770f8f1cd1fde6765b81d35aacaaf4ff98b82
Signed-off-by: Filip Tehlar <ftehlar@cisco.com>
|
|
* compiler -march= parameter is changed from native to corei7
so code is always genereted with instructions which are available
on the Nehalem microarchitecture (up to SSE4.2)
* compiler -mtune= parameter is added so code is optimized for
corei7-avx which equals to Sandy Bridge microarchitecture
* set of macros is added which allows run-time detection of available
cpu instructions (e.g. clib_cpu_supports_avx())
* set of macros is added which allows us to clone graph node funcitons
where cloned function is optmized for different microarchitecture
Those macros are using following attributes:
__attribute__((flatten))
__attribute__((target("arch=core-avx2)))
I.e. If applied to foo_node_fn() macro will generate cloned
functions foo_node_fn_avx2() and foo_node_fn_avx512() (future)
It will also generate function void * foo_node_fn_multiarch_select()
which detects available instruction set and returns pointer to the
best matching function clone.
Change-Id: I2dce0ac92a5ede95fcb56f47f3d1f3c4c040bac0
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
JIRA: VPP-66
Change-Id: I421529fa8eafe5268745a34a4fcd40156defcdf8
Signed-off-by: Matus Fabian <matfabia@cisco.com>
|
|
* adds support for VPP native PCI drivers using
standard uio_pci_generic kernel driver
* adds generic PCI interrupt callback
* splits code to generic PCI handling and linux specific
* adds new debug cli 'show pci [all]'
Change-Id: I447c2285e319e9725d70688c1b70c9dedda51fdc
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
In vnet/vnet/devices/dpdk/device.c, post 2 event data instead of
merging them into 1 u64.
Change-Id: I8b1d61b894279fb6eb57bb82a05affc14360e6b8
Signed-off-by: Christophe Fontaine <christophe.fontaine@qosmos.com>
|
|
On non numa aware platforms, the path "/sys/devices/system/node/node%u"
does not exist: the "free_hugepages" count is available in "/sys/kernel/mm/".
Change-Id: Id62018a2bd0937d267663699c620e89d02c26a72
Signed-off-by: Christophe Fontaine <christophe.fontaine@qosmos.com>
|
|
Change-Id: I1470993668fa8c740229149bbb227bd1e410d5ab
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Change-Id: I76359b621b2edc599cf2e9ee845d97293a5d46f7
Signed-off-by: Keith Burns (alagalah) <alagalah@gmail.com>
|