The VLIB_FRAME_NO_FREE_AFTER_DISPATCH flag is not preserved when cloning
next_frames, as a result VLIB_FRAME_FREE_AFTER_DISPATCH can
erroneously be set for a frame (see vlib_get_next_frame_internal())
Change-Id: Ice1d7ddcb807e1168aa0c157d9474be492d102c2
Signed-off-by: Nikhil P Rao <nikhil.rao@intel.com>
Fixing a previous commit to remove the commandline option and just
always display the DPDK physical memory on startup. Also remove
the DPDK args for the dbeug image.
Change-Id: I9336a2a26ebab84c3b50ae0610025c29da7995f2
Signed-off-by: Todd Foggoa <tfoggoa@cisco.com>
It turns out that unix_physmem_init(...) has been effectively disabled
for a very long time. The vnet library supplied a weak symbol override
for the vlib_app_physmem_init(...) which returned 1, meaning "do
nothing." When we switched libvnet.a -> libvnet.so, the symbol
override stopped working.
Presto: unix_physmem_init(...) romps all over the data set up by
vlib_buffer_pool_create(...), leading to ASSERT failures and/or bus
errors, but only when using worker threads. Even then, the failure
depended in some complicated way on library dynamic load order.
We should remove .../vlib/vlib/unix/physmem.c entirely once we're sure
we'll never want it back.
Change-Id: I27747edbeb0de88d2f2d8728f7f8eb3135e7f0cf
Signed-off-by: Dave Barach <dave@barachs.net>
Added new functions to change the MAC address and filter multicast MAC
addresses.
Change-Id: Iddf518e57dc889800a2f706fda51ee4e5c5142f2
Signed-off-by: Todd Foggoa (tfoggoa) <tfoggoa@cisco.com>
On admin up/down changes the TX and RX queues were getting locked
up for VNET_DPDK_PMD_VMXNET3 interfaces. The fix is to ensure the
interface is in admin down before touching the hardware.
Change-Id: Ia264d95355a860607bff7ada00d2be48f91e6dd9
Signed-off-by: Todd Foggoa (tfoggoa) <tfoggoa@cisco.com>
Export interface format functions to plugin and allow ability to
show a single hardware interface index.
Change-Id: If52fae2d63e97da91e1ac9a9a6fb73389b526ebc
Signed-off-by: Todd Foggoa (tfoggoa) <tfoggoa@cisco.com>
Vxlan interface link status corresponds to admin status.
When vxlan interfcae is set "admin-up" link status is "link-up" and if interface is set "admin-down" link status is "link-down" also.
Change-Id: Iaa85fa8dcd672519312cc7ee6a17df6f4498259d
Signed-off-by: Pavel Kotucek <pkotucek@cisco.com>
Note that compiling -O3 doesn't improve performance as of this
writing, might as well clean up warnings in any event.
Change-Id: Ic2f4982d12fbbf36f5324075183982731759dc94
Signed-off-by: Dave Barach <dave@barachs.net>
This avoids checking the buffer flags bit if tracing is not enabled.
Change-Id: I32e1a90b5fd10318254c611344488bc2a441c71e
Signed-off-by: Todd Foggoa (tfoggoa) <tfoggoa@cisco.com>
When it runs 'exec' command string "(nil)" is appended to the end of the
console output.
Change-Id: I7cdae78b2166829dd2160e92ed8181203eb491ed
Signed-off-by: Pavel Kotucek <pkotucek@cisco.com>
Standard Ubuntu PowerPC toolchain can be installed with:
apt-get install crossbuild-essential-powerpc
And code can be built with:
make PLATFORM=qppc TAG=qppc ...
Change-Id: I247355fc89fd61aedb8693156e84c454ca86eb75
Signed-off-by: Damjan Marion <damarion@cisco.com>
poll mode driver library. The bonded interfaces to be created on
VPP startup is specified in the dpdk section of startup.conf
or qn.conf, using DPDK EAL command. Following is an example of a
dpdk section white listing PCI addressses of 4 ethernet interfacess
to be under VPP control plus two bonded interface and the PCI addresses
of the slaves in each:
dpdk { socket-mem 1024,1024
dev 0000:0f:00.0 dev 0000:10:00.0 dev 0000:11:00.0 dev 0000:12:00.0
vdev eth_bond0,mode=2,slave=0000:0f:00.0,slave=0000:11:00.0,xmit_policy=l34
vdev eth_bond1,mode=2,slave=0000:10:00.0,slave=0000:12:00.0,xmit_policy=l34
}
Note that only balance XOR (mode 2) is supported and "xmit_policy=l34"
specifies to use layer 3 SIP/DIP and layer 4 Sport/Dport for load
balance. Using "xmit_policy=l2" for SMAC/DMAC or "xmit_policy=l23" for
SMAC/DMAC and SIP/DIP should also work.
Change-Id: Iaf6438686fa20cce893cb5a823b76e2886b4360b
Signed-off-by: John Lo <loj@cisco.com>
This patch adds live migration support to vhost interfaces, by supporting the VHOST_F_LOG_ALL feature. When qemu starts a migration, it will negotiate this feature, and provide a fd for a "dirty log" shared mem space. This log is a bitmap representing pages in the device memory.
Whenever we touch memory pointed by a "desc" vring, or modify a "used" vring, we log the corresponding page in the bitmap. This allows qemu to send the dirty page to the destination host.
See https://github.com/qemu/qemu/blob/master/docs/specs/vhost-user.txt, § "Live migration" for more details.
In addition to this, this code provides support for the VHOST_USER_F_PROTOCOL_FEATURES feature, and to VHOST_USER_{GET,SET}_PROTOCOL_FEATURES and VHOST_USER_SET_VRING_ENABLE messages, required for live migration.
Change-Id: I7577efce8bd67653218f4291af1d651de451e552
Signed-off-by: Yoann Desmouceaux <ydesmouc@cisco.com>
Turn of srp, mainly as an example of how to restructure a featurette
for selective disablement.
Change-Id: Id3364c58a8711b103939f4434adfa67177380f67
Signed-off-by: Dave Barach <dave@barachs.net>
If number of Qs enabled by the guest is less
than number of Qs exported to guest, then all
threads route packet to Q0 with the assumption
that guest is not interested in performance.
If all Qs are enabled, each thread queues packet
in their own TX-Q of the VHOST_USER interface,
boosting performance.
Change-Id: Ic24bb8c0505b11c7513aeecd21c9ec5da5f90138
Signed-off-by: Shesha Sreenivasamurthy <shesha@cisco.com>
Adding a patch to fix DPDK 2.2.0 enic PMD driver TX function.
The enic PMD driver send function uses a constant offset instead of relying on the data_off in the mbuf to find the start of the packet.
Change-Id: Ic4f3be83865367306785a57e2694e0ccfa295c7b
Signed-off-by: Yoann Desmouceaux <ydesmouc@cisco.com>
AtomicBoolean forces the enclosed boolean to be cache-aligned,
needlessly hitting a cacheline. Since we only flip state only once,
when the connection is severed, we can turn this into a volatile
boolean.
Doing that is costing us a synchronized close(), but that is perfectly
acceptable, as that is called only once.
Change-Id: I76fda3d3f65a5f800e7d3970b0b8fe99fb3e8b6d
Signed-off-by: Robert Varga <nite@hq.sk>
Generating a new MAC address everytime vpp reboots can quickly
become cumbersome (e.g. in test scenarios).
This commit allows vhost mac addresses to be configured
manually.
Change-Id: Ic6d16a2d14cdf8e49dc29041bbafde39d9fc8370
Signed-off-by: Pierre Pfister <ppfister@cisco.com>
Include node names and graph arcs. Prep work for uploading node
runtime data, so the latter can be reported in a comprehensible
manner.
Change-Id: I215b1f8cff244200c37c7e088f1f22229dc97eb6
Signed-off-by: Dave Barach <dave@barachs.net>
Adding a patch to fix DPDK enic PMD driver TX function.
The enic PMD driver send function uses a constant offset instead of relying on the data_off in the mbuf to find the start of the packet.
Change-Id: I74dbcce949f08e39070a369d992a90586af140a6
Signed-off-by: Yoann Desmouceaux <ydesmouc@cisco.com>
Validated following images used do not contain gdb, gdbserver:
puppetlabs/centos-7.0-64-nocm
puppetlabs/ubuntu-14.04-64-nocm
Note that centos7 does not include emacs
Change-Id: I091a64a7c667e081574763537de724f9feaae0dc
Signed-off-by: Keith Burns (alagalah) <alagalah@gmail.com>
This code provided inter-VM (2 cores per VM) throughput of
22Gbps using iperf through VPP (1 core) with 9k frames.
With the same setup and pktgen running on both sides, it
reached 5Mpps with no packets drop (Equivalent to before the patch).
During the tests the average vector length was about 1, which
likely means that VPP is not the bottleneck.
The patch also includes some generic functions for vlib buffers
allowing for chained buffer construction whether or not DPDK is enabled.
Change-Id: Icfd1803e84b2b4578f305ab730576211f6242d6a
Signed-off-by: Pierre Pfister <ppfister@cisco.com>
It also includes check to ensure that number of
per-cpu mheaps is not lower than number of cpus.
Change-Id: Ibc68b34dda130f922243f9ea15b03e44bbcac269
Signed-off-by: Damjan Marion <damarion@cisco.com>
vec_sort macro was using gcc proprietary nested functions that
require a executable stack and they are considered as unsafe.
Also, nested functions are not supported by other compilers.
vec_sort_with_function() should be used instead.
Change-Id: I05959da63d222ec71c090ba63420b427ce10c79b
Signed-off-by: Damjan Marion <damarion@cisco.com>
This patch alters behavior for dpdk rx packets. Depending on test
results, it may be necessary to extend the scheme to packets received
on non-dpdk paravirtualized interfaces, and packets originating in the
vpp stack itself.
Change-Id: I8444232a90ff176e7d6a688e36801174575251a1
Signed-off-by: Dave Barach <dave@barachs.net>
Maximum number of TX queues ca be defined by NIC driver
or configured manualy with 'dpdk { max-tx-queues X }'.
If system have more worker threads than TX queues they will
be shared between them. Before this change only one tx
queue was used in such cases.
Change-Id: Iab68170ab45fa6b9925fc4e79ccab9222f095e7e
Signed-off-by: Damjan Marion <damarion@cisco.com>
