Age | Commit message (Collapse) | Author | Files | Lines |
|
Too many prefetches within loop unrollings induce bottleneck and
performance degradation on some CPUs which have less cache line fill
buffers, e.g, Arm Cortex-A72.
Apply dual loop unrolling and tune prefetches manually to resolve
hot-spot with prefetch instructions.
It saves about 11.5% cycles with ip4_input node on Cortex-A72 CPUs.
Type: feature
Change-Id: I1ac9eb21061a804af2a414b420217fbcda3689c9
Signed-off-by: Lijian Zhang <Lijian.Zhang@arm.com>
|
|
Too many prefetches within loop unrollings induce bottleneck and
performance degradation on some CPUs which have less cache line fill
buffers, e.g, Arm Cortex-A72.
Apply dual loop unrolling and tune prefetches manually to remove
hot-spot with prefetch instructions, to get throughput improvement.
It brings about 7% throughput improvement and saves 28% clocks with
ip4_rewrite nodes on Cortex-A72 CPUs.
Type: feature
Change-Id: I0d35ef19faccbd7a5a4647f50bc369bfcb01a20d
Signed-off-by: Lijian Zhang <Lijian.Zhang@arm.com>
|
|
Use consistent API types.
Type: fix
Signed-off-by: Jakub Grajciar <jgrajcia@cisco.com>
Change-Id: Idbba4ab6a412b75338e3149e51476693f0862f16
Signed-off-by: Jakub Grajciar <jgrajcia@cisco.com>
|
|
Type: feature
Change-Id: I201155b1b92cf7e57310af726879dab039090582
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Type: feature
Change-Id: If68d07fbe8c6f7fffd2f93c7e854367082927e4f
Signed-off-by: Simon Zhang <yuwei1.zhang@intel.com>
|
|
Separate debug CLI arg parsing from the underlying action
function. Fixes a number of subtle ordering dependencies, and will
allow us to add a binary API to control the feature at some point in
the future.
Type: refactor
Ticket: VPP-1770
Signed-off-by: Dave Barach <dave@barachs.net>
Change-Id: Id0dbeda06dad20e756c941c691e2088ce3c50ec7
|
|
Type:fix
Change-Id: I173f45bf3d90e6979675b2ac812a969539f02147
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Type: feature
This shouldn't happen unless connections are removed after buffers were
enqueued to tcp-output and before tcp-output runs. For instance, packet
is enqueued to tcp output and cleanup is called for connection.
Change-Id: Ib7dd82ffa6cfb21ff5068aba010e0a3497eeea13
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
When checksumming chained buffers with odd lengths: insert a
NULL byte, or the calculation fails.
Type: fix
Signed-off-by: Dave Barach <dave@barachs.net>
Signed-off-by: John Lo <loj@cisco.com>
Change-Id: I380f7c42897bdb28c8c29aa1c4cdaaa849cc9ecc
|
|
Not all interfaces have the same characteristics within the bonding group.
For active-backup mode, we should do our best to select the slave that
performs the best as the primary slave. We already did that by preferring
the slave that is local numa. Sometimes, this is not enough. For example,
when all are local numas, the selection is arbitrary. Some slave interfaces
may have higher speed or better qos than the others. But this is hard to
infer.
One rule does not fit all. So we let the operator to optionally specify the
weight for each slave interface. Our primary slave selection rule is now
1. biggest weight
2. is local numa
3. current primary slave (to avoid churn)
4. lowest sw_if_index (for deterministic behavior)
This selection rule only applies to active-backup mode which only one slave
is used for forwarding traffic until it becomes unreachable. At that time,
the next "best" slave candidate is automatically promoted. The slaves are
sorted according to the preference rule when they are up. So there is no need
to find the next best candidate when the primary slave goes down.
Another good thing about this rule is when the down slave comes back up, it
is selected as the primary slave again unless there is indeed a "better"
slave than this down slave that were added during that period.
To set the weight for the slave interface, do this after the interface is
enslaved
set interface bond <interface-name> weight <value>
Type: feature
Signed-off-by: Steven Luong <sluong@cisco.com>
Change-Id: I59ced6d20ce1dec532e667dbe1afd1b4243e04f9
|
|
Type: fix
Ticket: VPP-1766
revert e4ac48e792f4eebfce296cfde844ee73b1abd62f
Change-Id: I03feea4008a47859d570ad8d1d08ff3f30d139ef
Signed-off-by: Mohsin Kazmi <sykazmi@cisco.com>
(cherry picked from commit 623a1b7053424b539a51faf866ab839d3da3f45b)
|
|
Use consistent API types.
Type: fix
Signed-off-by: Jakub Grajciar <jgrajcia@cisco.com>
Change-Id: I4b91ebe8aadbec8b95716ed1cde4b0a13eee3a3f
Signed-off-by: Jakub Grajciar <jgrajcia@cisco.com>
|
|
Add /if/lacp/<bond-sw_if_index>/<slave-sw_if_index>/state
<bond-sw_if_index> is a vector of the bond sw_if_index
<slave-sw_if_index> is a vector of the slave sw_if_index
Content is the integer value of the lacp actor state. The state is actually
a bitfield as described in the lacp protocol spec.
Type: feature
Signed-off-by: Steven Luong <sluong@cisco.com>
Change-Id: Ic6eca8ce2a1acd2d858e4e50b7eac1d000ea08e5
Signed-off-by: Ole Troan <ot@cisco.com>
|
|
Type: feature
Allow session cli filtering based on thread index, transport protocol,
session state and range of session pool indices. For instance
show session thread 1 proto tcp state ready range 0 20 verbose
Shows the session ids for the first 20 tcp sessions in thread 1 that are
in ready state.
To avoid excessive output that could reasult in the worker barrier being
held by the main thread for long periods of time, the session cli will
only output:
- session ids (verbose == 1) for a maximum of 50 sessions / worker
- verbose > 1 details for a maximum of 10 sessions
Change-Id: I2cfb351b548e2e0a1d5b4345810be613e2917d17
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Type: refactor
Change-Id: I15b10a22d0d0b83075a0eef5ef8c09cf76989866
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
In eth_input_process_frame(), destination MAC addresses are
compared to the interface hardware address in a loop. Move
this to a separate inline function to facilitate making changes
to the filtering logic more cleanly.
Type: refactor
Change-Id: I0978f01667e78af5214dbbc9ba223f5b84ce6b7e
Signed-off-by: Matthew Smith <mgsmith@netgate.com>
|
|
Type: fix
Change-Id: I1dbc822061f43fcc25dfa5d1728f5c7575139f45
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
In eth_input_process_frame(), a loop which checks the destination
MAC address of received packets had a different condition for avx2
than it did for the non-avx2 version of the loop. It could result
in one unnecessary execution of the loop body after all packets
had been processed.
Type: fix
Fixes: 8d6f34e2b1cbfde5702e88686631527d5e1e10a0
Change-Id: Ib673f45917a0dea461987fdc8f0ca318b749fb1a
Signed-off-by: Matthew Smith <mgsmith@netgate.com>
|
|
Enforce that variable length fields are the last element of API messages.
Add a 'fixed' version of string type, since dealing with
multiple variable length strings turned out too painful
for the C language bindings.
The string type is now:
{
string name[64]; // NUL terminated C-string. Essentially decays to u8 name[64]
string name[]; // Variable length string with embedded len field (vl_api_string_t)
};
The latter notation could be made available to other types as well.
e.g.
{
vl_api_address_t addresses[];
}
instead of
{
u32 n_addr;
vl_api_address_t addresses[n_addr];
};
Type: fix
Change-Id: I18fa17ef47227633752ab50453e8d20a652a9f9b
Signed-off-by: Ole Troan <ot@cisco.com>
|
|
The VPP code tries to set all userspace memory in the table via IOCTL
to VHOST_SET_MEM_TABLE. But on aarch64, the userspace address range is
larger (48 bits) than that on x86 (47 bits). Below is an segment from
/proc/[vpp]/maps.
fffb41200000-fffb43a00000 rw-s 00000000 00:0e 532232
/anon_hugepage (deleted)
Instead of setting all userspace memory space to vhost-net, will only set
the address space reserved by pmalloc module during initialization.
Type: fix
Change-Id: I91cb35e990869b42094cf2cd0512593733d33677
Signed-off-by: Lijian Zhang <Lijian.Zhang@arm.com>
Reviewed-by: Steve Capper <Steve.Capper@arm.com>
|
|
Type: feature
Change-Id: I3e97e05a31806afb6b2e84ecf05fb96d285db92e
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
vlib_increment_combined_counter takes sw_if_index, not hw_if_index. Using
hw_if_index may work as long as there is no subinterface created to cause
hw_if_index and sw_if_index to differ.
Type: fix
Signed-off-by: Steven Luong <sluong@cisco.com>
Change-Id: I6db042186eeeacf32250f7ef261af8cd6f5ce56e
|
|
Credits to ray.kinsella@intel.com who spotted the issue and identified
root cause.
Type: fix
Change-Id: I4afe74c47769484309f6aebca2de56ad32c8041f
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Type: fix
Change-Id: Iff4a13da654dbdf7f08bb6d0743b466a1e6be01b
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Type: fix
in case the app requests another dequeue notification, it should not be cleared
Change-Id: I7e94e8e30ef098f87f5e7f4880af276970ee5b37
Signed-off-by: Vladimir Kropylev <vladimir.kropylev@enea.com>
|
|
Per interface, next hop graph node can be customized
with vnet_hw_interface_rx_redirect_to_node function,
but it doesn't work well for af-packet type interface.
In current implementation, with function
af_packet_set_interface_next_node invoked next hop graph
node index can be set to apif->per_interface_next_index,
but it's not set to next0 properly for packet processing
in af_packet_device_input_fn.
Type: fix
Signed-off-by: Michael Yu <michael.a.yu@nokia-sbell.com>
Change-Id: I8e132ddd1c3c01b6f476de78546d4a9389b3ff87
Signed-off-by: Michael Yu <michael.a.yu@nokia-sbell.com>
|
|
Type: feature
Change-Id: I7c5684a9f88e56fd62d83a59c14c4e4a6f096b89
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Type: fix
Use different mss for v4 and v6 packets.
Change-Id: Ifcd322c1fd74f749a42bae671dd6cfed8c32b150
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Type: feature
Distinguish between rcv window errors and errors after we advertised a
zero rcv window, i.e., potential window probes.
Change-Id: I6cb453c7aaae456c0a05a8328cfaa55eaca10bf7
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Type: fix
show session rules tcp
vpp/src/vnet/session/application_namespace.c:34 (app_namespace_get)
assertion `! pool_is_free (app_namespace_pool, _e)' fails
Thread 1 "vpp_main" received signal SIGABRT, Aborted.
Signed-off-by: Guanghua Zhang <ghzhang@fiberhome.com>
Change-Id: I3430339a598a92d0a38826471adeb4547b0b7df4
|
|
Type: refactor
Signed-off-by: Simon Zhang <yuwei1.zhang@intel.com>
Change-Id: I8674ff5f6f6336b256b7df8187afbb36ddef71fb
|
|
Type: fix
Change-Id: Ia362ad821db1fd506e973e1844cc3ec74703cc17
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Use consistent API types.
Type: fix
Change-Id: I416171c729193e64a68e85a3f1f25d03a9a559d1
Signed-off-by: Jakub Grajciar <jgrajcia@cisco.com>
|
|
Type:fix
Change-Id: If4fce6bd9b06ed545d72df847afd6a74cdefe042
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Type: feature
with detail option prints all details for all SAs
Change-Id: Ic3c423c085dfc849cf9c3e18a6f624b82150d961
Signed-off-by: Neale Ranns <nranns@cisco.com>
|
|
dpdk bonding code was removed in 19.08. However, there are still references
to VNET_SW_INTERFACE_FLAG_BOND_SLAVE which was set by the already removed
code.
Type: fix
Signed-off-by: Steven Luong <sluong@cisco.com>
Change-Id: I949a7281e6273f2733dd7532cc4a3bb4f3ce30de
|
|
Type:refactor
Moves connect, disconnect, bind, unbind and app detach to message
queue from binary api. Simplifies app/vcl interaction with the session
layer since all session control messages are now handled over the mq.
Add/del segment messages require internal C api changes which affect all
builtin applications. They'll be moved in a different patch and might
not be back portable to 19.08.
Change-Id: I93f6d18e551b024effa75d47f5ff25f23ba8aff5
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Type: fix
Ticket: VPP-1756
the block-size was set to 0 resulting in incorrect placement of the ESP
footer.
add tests for NULL encrypt + integ.
Change-Id: I8ab3afda8e68f9ff649540cba3f2cac68f12bbba
Signed-off-by: Neale Ranns <nranns@cisco.com>
|
|
Type: refactor
Change-Id: I30cb324006c3c6ab88d01c044cf80fab3a34f13a
Signed-off-by: Klement Sekera <ksekera@cisco.com>
|
|
Type: fix
Ticket: VPP-1751
Change-Id: I5ffb078492adc97374290de404f2ec0102b75184
Signed-off-by: Mohsin Kazmi <sykazmi@cisco.com>
|
|
The as_u64[0] in the result was not correctly filled in for
longer prefix lengths.
Type: fix
Fixes: 1c7104514c
Change-Id: I871772c618475396d1c1c4c699ff77e35097f07e
Signed-off-by: Andrew Yourtchenko <ayourtch@gmail.com>
|
|
Type: fix
Change-Id: I3b4cb8fc17387693d6ad6cdef004ca99cd56bc23
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Type: refactor
Signed-off-by: Dave Barach <dave@barachs.net>
Change-Id: I823c2cd307a4613653a2d20f564dda062d4da31b
|
|
Signed-off-by: Klement Sekera <ksekera@cisco.com>
Change-Id: Idd64b0b3eedb7179f6d316b66cea8160ad577990
Type: fix
Fixes: 630ab5846bceddf8d663e9f488a2dc0378949827
Signed-off-by: Klement Sekera <ksekera@cisco.com>
|
|
Can we get a reply macro that would avoid such copypaste errors?
Type: fix
Change-Id: I753efb1d82fced668c27f2e44b0318bfd31c0a23
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
Type: feature
Add option to always compile in debugging infra. Debug verbosity can be
controlled via configuration updatable via cli. Compile time inclusion
and configuration of event logging verbosity is still available in order
to minimize event logging impact.
Change-Id: I9d946efe9fa4204fdace1adb1d6588c97b5ae758
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
Type: test
Signed-off-by: Dave Barach <dave@barachs.net>
Change-Id: I88e93b510d8d4f7b59f3e630539dc7e0264afa47
|
|
Type:fix
Change-Id: Ib5c33f0bdb3a8b1c2585135fcc8c07b151e47df3
Signed-off-by: Florin Coras <fcoras@cisco.com>
|
|
vppapigen has remapped legacy to typedefs behind the scenes
for some time now.
- update .api files to use new style typedefs.
- issue error on 'typeonly define' in .api files
- remove unneeded macros redefining vl_noop_handler
Type: refactor
Change-Id: I7a8c4a6dafacee6a131f95cd0e9b03a8c60dea8b
Signed-off-by: Paul Vinciguerra <pvinci@vinciconsulting.com>
|
|
Note which worker received fragment with offset zero and use this worker
to send out the reassembled packet.
Type: fix
Change-Id: I1d3cee16788db3b230682525239c0100d51dc380
Signed-off-by: Klement Sekera <ksekera@cisco.com>
|