dpdk.git
7 years agopdump: add new library for packet capture
Reshma Pattan [Wed, 15 Jun 2016 14:06:22 +0000 (15:06 +0100)]
pdump: add new library for packet capture

The librte_pdump library provides a framework for
packet capturing in dpdk. The library provides set of
APIs to initialize the packet capture framework, to
enable or disable the packet capture, and to uninitialize
it.

The librte_pdump library works on a client/server model.
The server is responsible for enabling or disabling the
packet capture and the clients are responsible
for requesting the enabling or disabling of the packet
capture.

Enabling APIs are supported with port, queue, ring and
mempool parameters. Applications should pass on this information
to get the packets from the dpdk ports.

For enabling requests from applications, library creates the client
request containing the mempool, ring, port and queue information and
sends the request to the server. After receiving the request, server
registers the Rx and Tx callbacks for all the port and queues.
After the callbacks registration, registered callbacks will get the
Rx and Tx packets. Packets then will be copied to the new mbufs that
are allocated from the user passed mempool. These new mbufs then will
be enqueued to the application passed ring. Applications need to dequeue
the mbufs from the rings and direct them to the devices like
pcap vdev for viewing the packets outside of the dpdk
using the packet capture tools.

For disabling requests, library creates the client request containing
the port and queue information and sends the request to the server.
After receiving the request, server removes the Rx and Tx callback
for all the port and queues.

Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
7 years agoethdev: get port by name
Reshma Pattan [Wed, 15 Jun 2016 14:06:21 +0000 (15:06 +0100)]
ethdev: get port by name

Converted rte_eth_dev_get_port_by_name to a public API.
Converted rte_eth_dev_get_name_by_port to a public API.

The librte_pdump library provides the APIs to enable or disable the
packet capture either using the port id or pci address or device name.
So pdump library need to do a mapping from name to port and port to name
internally to validate the device name and register the Rx and Tx
callbacks for the mapped ports. So these APIs are made public for the
pdump library for doing the mentioned mappings.

Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
7 years agoethdev: get queues numbers
Reshma Pattan [Wed, 15 Jun 2016 14:06:20 +0000 (15:06 +0100)]
ethdev: get queues numbers

The new fields nb_rx_queues and nb_tx_queues are added to the
rte_eth_dev_info structure.
Changes to API rte_eth_dev_info_get() are done to update these new fields
to the rte_eth_dev_info object.
Release notes is updated with the changes.

The librte_pdump library needs to register Rx and Tx callbacks for all
the nb_rx_queues and nb_tx_queues, when application wants to capture the
packets on all the software configured number of Rx and Tx queues of the
device. So far there is no support to get nb_rx_queues and nb_tx_queues
information from the ethdev library. Hence these changes are introduced.

Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
7 years agoethdev: insert Rx callback as head of list
Reshma Pattan [Wed, 15 Jun 2016 14:06:19 +0000 (15:06 +0100)]
ethdev: insert Rx callback as head of list

Added new public api rte_eth_add_first_rx_callback to add given
callback as head of the list.

The librte_pdump library should display Rx packets of the
NIC even before they are being processed by other callbacks
of the application (because other callbacks of the application
may change the packet data as part of the processing).
So packet capturing framework should register a callback at the
head of the Rx callback list so that callback always gets called
first before any other callbacks of the applications. Hence this API
is introduced.

Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
7 years agoethdev: protect Rx/Tx callback change with locks
Reshma Pattan [Wed, 15 Jun 2016 14:06:18 +0000 (15:06 +0100)]
ethdev: protect Rx/Tx callback change with locks

Added spinlocks around add/remove logic of Rx and Tx callbacks
to avoid corruption of callback lists in multithreaded context.

Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
7 years agoexamples/l2fwd-keepalive: add IPC liveness reporting
Remy Horton [Wed, 15 Jun 2016 15:25:51 +0000 (16:25 +0100)]
examples/l2fwd-keepalive: add IPC liveness reporting

Changes the l2fwd keepalive example to show how the new keepalive
enhancements can be used to relay core state to an external process.

Signed-off-by: Remy Horton <remy.horton@intel.com>
7 years agokeepalive: add liveness callback
Remy Horton [Wed, 15 Jun 2016 15:25:50 +0000 (16:25 +0100)]
keepalive: add liveness callback

Adds and documents new callbacks that allow transitions to core
states other than dead to be reported to applications.

Signed-off-by: Remy Horton <remy.horton@intel.com>
7 years agokeepalive: export states
Remy Horton [Wed, 15 Jun 2016 15:25:49 +0000 (16:25 +0100)]
keepalive: export states

Changes the keepalive state from an anonymous enum to a declared one
which is externally visible, so that keepalive enum values can be
used by applications.

Signed-off-by: Remy Horton <remy.horton@intel.com>
7 years agoethdev: remove name from extended statistic fetch
Remy Horton [Wed, 15 Jun 2016 15:25:33 +0000 (16:25 +0100)]
ethdev: remove name from extended statistic fetch

The current extended ethernet statistics fetching involve doing several
string operations, which causes performance issues if there are lots of
statistics and/or network interfaces. This patch changes the test-pmd
and proc_info applications to use the new xstats API, and removes
deprecated code associated with the old API.

Signed-off-by: Remy Horton <remy.horton@intel.com>
7 years agonet/virtio: fetch extended statistics with integer ids
Remy Horton [Wed, 15 Jun 2016 15:25:32 +0000 (16:25 +0100)]
net/virtio: fetch extended statistics with integer ids

The current extended ethernet statistics fetching involve doing several
string operations, which causes performance issues if there are lots of
statistics and/or network interfaces. This patch changes the virtio driver
to use the new API that seperates name string and value queries.

Signed-off-by: Remy Horton <remy.horton@intel.com>
7 years agonet/i40e: fetch extended statistics with integer ids
Remy Horton [Wed, 15 Jun 2016 15:25:31 +0000 (16:25 +0100)]
net/i40e: fetch extended statistics with integer ids

The current extended ethernet statistics fetching involve doing several
string operations, which causes performance issues if there are lots of
statistics and/or network interfaces. This patch changes the i40e driver
to use the new API that seperates name string and value queries.

Signed-off-by: Remy Horton <remy.horton@intel.com>
7 years agonet/fm10k: fetch extended statistics with integer ids
Remy Horton [Wed, 15 Jun 2016 15:25:30 +0000 (16:25 +0100)]
net/fm10k: fetch extended statistics with integer ids

The current extended ethernet statistics fetching involve doing several
string operations, which causes performance issues if there are lots of
statistics and/or network interfaces. This patch changes the fm10k driver
to use the new API that seperates name string and value queries.

Signed-off-by: Remy Horton <remy.horton@intel.com>
7 years agonet/igb: fetch extended statistics with integer ids
Remy Horton [Wed, 15 Jun 2016 15:25:29 +0000 (16:25 +0100)]
net/igb: fetch extended statistics with integer ids

The current extended ethernet statistics fetching involve doing several
string operations, which causes performance issues if there are lots of
statistics and/or network interfaces. This patch changes the e1000 driver
to use the new API that seperates name string and value queries.

Signed-off-by: Remy Horton <remy.horton@intel.com>
7 years agonet/ixgbe: fetch extended statistics with integer ids
Remy Horton [Wed, 15 Jun 2016 15:25:28 +0000 (16:25 +0100)]
net/ixgbe: fetch extended statistics with integer ids

The current extended ethernet statistics fetching involve doing several
string operations, which causes performance issues if there are lots of
statistics and/or network interfaces. This patch changes the ixgbe driver
to use the new API that seperates name string and value queries.

Signed-off-by: Remy Horton <remy.horton@intel.com>
7 years agoethdev: fetch extended statistics with integer ids
Remy Horton [Wed, 15 Jun 2016 15:25:27 +0000 (16:25 +0100)]
ethdev: fetch extended statistics with integer ids

The current extended ethernet statistics fetching involve doing several
string operations, which causes performance issues if there are lots of
statistics and/or network interfaces. This patch changes the xstats
functions to instead use a numeric identifier rather than a string, and
adds the ability to retrieve identifier-to-string mappings.

Signed-off-by: Remy Horton <remy.horton@intel.com>
7 years agoapp/testpmd: print RSS hash update error
Slawomir Mrozowicz [Wed, 15 Jun 2016 10:33:42 +0000 (12:33 +0200)]
app/testpmd: print RSS hash update error

Calling rte_eth_dev_rss_hash_update without checking return value.
Fixed by handle return value and print out error status.

Coverity issue: 119251
Fixes: ce8d561418d4 ("app/testpmd: add port configuration settings")

Signed-off-by: Slawomir Mrozowicz <slawomirx.mrozowicz@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
7 years agoapp/testpmd: show topology at forwarding start
Zhihong Wang [Tue, 14 Jun 2016 23:08:06 +0000 (19:08 -0400)]
app/testpmd: show topology at forwarding start

This patch show topology at forwarding start.

"show config fwd" also does this, but showing it directly can reduce the
possibility of misconfiguration.

Currently fwd_config_display() calls fwd_config_setup(), this misleading
behavior will be fixed in other patches.

Signed-off-by: Zhihong Wang <zhihong.wang@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
7 years agoapp/testpmd: handle all Rx queues in RSS setup
Zhihong Wang [Tue, 14 Jun 2016 23:08:05 +0000 (19:08 -0400)]
app/testpmd: handle all Rx queues in RSS setup

This patch removes constraints in rxq handling when multiqueue is enabled
to handle all the rxqs.

Current testpmd forces a dedicated core for each rxq, some rxqs may be
ignored when core number is less than rxq number, and that causes confusion
and inconvenience.

One example: an engineer was doing multiqueue test, there're 2
ports in guest each with 4 queues, and testpmd was used as the forwarding
engine in guest, as usual he used 1 core for forwarding, as a results he
only saw traffic from port 0 queue 0 to port 1 queue 0, then a lot of
emails and quite some time are spent to root cause it, and of course it's
caused by this unreasonable testpmd behavior.

Moreover, even if we understand this behavior, if we want to test the
above case, we still need 8 cores for a single guest to poll all the
rxqs, obviously this is too expensive.

We met quite a lot cases like this, one recent example:
http://openvswitch.org/pipermail/dev/2016-June/072110.html

Signed-off-by: Zhihong Wang <zhihong.wang@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
7 years agoapp/testpmd: show throughput in port stats
Zhihong Wang [Tue, 14 Jun 2016 23:08:04 +0000 (19:08 -0400)]
app/testpmd: show throughput in port stats

This patch adds throughput numbers (in the period since last use of this
command) in port statistics display for "show port stats (port_id|all)".

Signed-off-by: Zhihong Wang <zhihong.wang@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
7 years agoapp/testpmd: add tx_first burst number option
Zhihong Wang [Tue, 14 Jun 2016 23:08:03 +0000 (19:08 -0400)]
app/testpmd: add tx_first burst number option

This patch enables configurable tx_first burst number.

Use "start tx_first (burst_num)" to specify how many bursts of packets to
be sent before forwarding start, or "start tx_first" like before for the
default 1 burst send.

Signed-off-by: Zhihong Wang <zhihong.wang@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
7 years agoapp/testpmd: add retry option
Zhihong Wang [Tue, 14 Jun 2016 23:08:02 +0000 (19:08 -0400)]
app/testpmd: add retry option

This patch adds retry option in testpmd to prevent most packet losses.
It can be enabled by "set fwd <mode> retry". All modes except rxonly
support this option.

Adding retry mechanism expands test case coverage to support scenarios
where packet loss affects test results.

Signed-off-by: Zhihong Wang <zhihong.wang@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
7 years agoapp/testpmd: stop forwarding on exit
Bernard Iremonger [Tue, 14 Jun 2016 15:35:42 +0000 (16:35 +0100)]
app/testpmd: stop forwarding on exit

Stop forwarding on exit whether all ports are started or not.

Fixes: ce8d561418d4 ("app/testpmd: add port configuration settings")

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
7 years agoapp/testpmd: check for valid mbuf pool
Bernard Iremonger [Tue, 14 Jun 2016 15:35:41 +0000 (16:35 +0100)]
app/testpmd: check for valid mbuf pool

Fixes: b6ea6408fbc7 ("ethdev: store numa_node per device")

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
7 years agoapp/testpmd: check for valid socket id when attaching port
Bernard Iremonger [Tue, 14 Jun 2016 15:35:40 +0000 (16:35 +0100)]
app/testpmd: check for valid socket id when attaching port

Fixes: edab33b1c01d ("app/testpmd: support port hotplug")

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
7 years agoapp/testpmd: move forward streams initialisation
Bernard Iremonger [Tue, 14 Jun 2016 15:35:39 +0000 (16:35 +0100)]
app/testpmd: move forward streams initialisation

Move call to init_fwd_streams from start_port function
to start_packet_forwarding function.

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
7 years agoapp/testpmd: separate forward config setup from display
Bernard Iremonger [Tue, 14 Jun 2016 15:35:37 +0000 (16:35 +0100)]
app/testpmd: separate forward config setup from display

Add call to fwd_config_setup to init_config.
Add call to fwd_config_setup for corelist, coremask and nbcore setup.
Add call to fwd_config_setup for portlist, portmask and nbport setup.
Add call to fwd_config_setup for rxq, txq, rxd and txd setup.
Remove fwd_config_setup from fwd_config_display.
Call the pkt_fwd_config_display function directly,
and remove the fwd_config_display wrapper function.

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
7 years agoapp/testpmd: check stopping port is not in bonding
Bernard Iremonger [Tue, 14 Jun 2016 15:35:38 +0000 (16:35 +0100)]
app/testpmd: check stopping port is not in bonding

Add new function port_is_bonding_slave
Use this function in stop_port and close_port functions.

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
7 years agoapp/testpmd: check stopping port is not forwarding
Bernard Iremonger [Tue, 14 Jun 2016 15:35:36 +0000 (16:35 +0100)]
app/testpmd: check stopping port is not forwarding

Add calls to port_is_forwarding function in stop_port and
close_port functions to check that port is not forwarding.

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
7 years agoapp/testpmd: do not update forwarding config
Bernard Iremonger [Tue, 14 Jun 2016 15:35:35 +0000 (16:35 +0100)]
app/testpmd: do not update forwarding config

Do not update forwarding configuration when attaching or detaching a port
Remove checks on test_done variable.
Remove code to update forwarding configuration.

Fixes: edab33b1c01d ("app/testpmd: support port hotplug")

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
7 years agoconfig: enable virtio for ppc64
Olivier Matz [Tue, 17 May 2016 09:59:52 +0000 (11:59 +0200)]
config: enable virtio for ppc64

Now that virtio pmd is supported on ppc, enable it.

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
7 years agovirtio: fix PCI accesses for ppc64 in legacy mode
David Marchand [Tue, 17 May 2016 09:59:51 +0000 (11:59 +0200)]
virtio: fix PCI accesses for ppc64 in legacy mode

Although ppc supports both endianesses, qemu supposes that the cpu is
big endian and enforces this for the virtio-net stuff.

Fix PCI accesses in legacy mode. Only ppc64le is supported at the moment.

Signed-off-by: David Marchand <david.marchand@6wind.com>
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
7 years agopci: mmap ioports on non-x86 Linux
Olivier Matz [Wed, 18 May 2016 16:27:30 +0000 (18:27 +0200)]
pci: mmap ioports on non-x86 Linux

On PPC64, the ioports are mapped in memory. Implement the missing part
of ioport API for PPC64 when using uio. This may also work on other
architectures but it has not been tested.

Signed-off-by: David Marchand <david.marchand@6wind.com>
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
7 years agopci: split function parsing resources in sysfs
Olivier Matz [Tue, 17 May 2016 09:59:49 +0000 (11:59 +0200)]
pci: split function parsing resources in sysfs

Split pci_parse_sysfs_resource() and introduce
pci_parse_one_sysfs_resource() that parses one line of sysfs resource
file.

This new function will be exported and used in next commits when
mapping the ioports resources.

No functional change.

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
7 years agopci: remove invalid comment
Olivier Matz [Tue, 17 May 2016 09:59:48 +0000 (11:59 +0200)]
pci: remove invalid comment

In a previous commit, the file used to map the PCI resources changed
from "/dev/uio<x>" to "/sys/bus/pci/devices/<busaddr>/resource", making
the comment wrong. Remove it.

Fixes: 9e67561acd1a ("eal/linux: mmap uio resources using resourceX files")

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
7 years agoeal/linux: only call iopl on x86
Olivier Matz [Tue, 17 May 2016 09:59:47 +0000 (11:59 +0200)]
eal/linux: only call iopl on x86

From iopl(2) man page: "This call is mostly for the x86 architecture. On
many other architectures it does not exist or will always return an
error".

This patch removes the call to iopl() in rte_eal_iopl_init() for
architectures other than x86, and always return 0 (success). This was
already done for ARM in
commit 0291476ae364 ("eal/linux: never check iopl for arm")

Next patches will introduce the support of memory mapped IO resources
for architectures != x86.

On BSD, there is nothing to do as open("/dev/io") already does the
proper thing. See man IO(4).

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
7 years agopci: fix typos in ioport doxygen comments
Olivier Matz [Tue, 17 May 2016 09:59:46 +0000 (11:59 +0200)]
pci: fix typos in ioport doxygen comments

Fix some typos and add missing comments related to ioports API in
rte_pci.h.

Fixes: 756ce64b1 ("eal: introduce PCI ioport API")

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
7 years agombuf: add new Rx flags for stripped VLAN
Olivier Matz [Wed, 15 Jun 2016 11:48:07 +0000 (13:48 +0200)]
mbuf: add new Rx flags for stripped VLAN

The behavior of PKT_RX_VLAN_PKT was not very well defined, resulting in
PMDs not advertising the same flags in similar conditions.

Following discussion in [1], introduce 2 new flags PKT_RX_VLAN_STRIPPED
and PKT_RX_QINQ_STRIPPED that are better defined:

  PKT_RX_VLAN_STRIPPED: a vlan has been stripped by the hardware and its
  tci is saved in mbuf->vlan_tci. This can only happen if vlan stripping
  is enabled in the RX configuration of the PMD.

For now, the old flag PKT_RX_VLAN_PKT is kept but marked as deprecated.
It should be removed from applications and PMDs in a future revision.

This patch also updates the drivers. For PKT_RX_VLAN_PKT:

- e1000, enic, i40e, mlx5, nfp, vmxnet3: done, PKT_RX_VLAN_PKT already
  had the same meaning than PKT_RX_VLAN_STRIPPED, minor update is
  required.
- fm10k: done, PKT_RX_VLAN_PKT already had the same meaning than
  PKT_RX_VLAN_STRIPPED, and vlan stripping is always enabled on fm10k.
- ixgbe: modification done (vector and normal), the old flag was set
  when a vlan was recognized, even if vlan stripping was disabled.
- the other drivers do not support vlan stripping.

For PKT_RX_QINQ_PKT, it was only supported on i40e, and the behavior was
already correct, so we can reuse the same bit value for
PKT_RX_QINQ_STRIPPED.

[1] http://dpdk.org/ml/archives/dev/2016-April/037837.html,

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
7 years agoethdev: clarify the origin of mbufs
Olivier Matz [Mon, 13 Jun 2016 13:45:28 +0000 (15:45 +0200)]
ethdev: clarify the origin of mbufs

Following the discussion on dpdk-users [1], enhance the API
documentation of rte_eth_tx_burst() to specify that the mbufs have to be
allocated from a pool.

[1] http://dpdk.org/ml/archives/users/2016-June/000618.html

Reported-by: Xiaoban Wu <xiaoban_wu@student.uml.edu>
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
7 years agoexamples: fix duplicated lpm6 name
Olivier Matz [Mon, 13 Jun 2016 09:43:03 +0000 (11:43 +0200)]
examples: fix duplicated lpm6 name

When starting the ip_fragmentation or ip_reassembly example
on several sockets, it fails.

The name of the lpm6 table is the same on every socket,
resulting in a table creation failure (-EEXIST). The failure
appeared after:
commit f82f705b635d ("lpm: fix allocation of an existing object")

Indeed, before this commit the returned value when the existing
table, which was probably a bug in that case: one table for 2
sockets for lpm6, and one per socket for lpm.

Fixes: 74de12b7b63a ("examples/ip_fragmentation: overhaul")
Fixes: b84fb4cb88ff ("examples/ip_reassembly: overhaul")

Reported-by: Heng Ding <hengx.ding@intel.com>
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
7 years agoeal/x86: improve memcpy performance
Zhihong Wang [Wed, 25 May 2016 01:23:03 +0000 (21:23 -0400)]
eal/x86: improve memcpy performance

This patch fixes rte_memcpy performance in Haswell and Broadwell for
vhost when copy size larger than 256 bytes.

It is observed that for large copies like 1024/1518 ones, rte_memcpy
suffers high ratio of store buffer full issue which causes pipeline
to stall in scenarios like vhost enqueue. This can be alleviated by
adjusting instruction layout. Note that this issue may not be visible
in micro test.

How to reproduce?

PHY-VM-PHY using vhost/virtio or vhost/virtio loop back, with large
packets like 1024/1518 bytes ones. Make sure packet generation rate
is not the bottleneck if PHY-VM-PHY is used.

Test report: http://dpdk.org/ml/archives/dev/2016-May/039716.html

Signed-off-by: Zhihong Wang <zhihong.wang@intel.com>
Tested-by: Qian Xu <qian.q.xu@intel.com>
7 years agomempool: fix local cache initialization
Sergio Gonzalez Monroy [Thu, 9 Jun 2016 08:19:47 +0000 (09:19 +0100)]
mempool: fix local cache initialization

The mempool local cache was not initialized properly leading to
undefined behavior in cases where the allocated memory was used
previously and left with data.

Fixes: 213af31e0960 ("mempool: reduce structure size if no cache needed")

Signed-off-by: Sergio Gonzalez Monroy <sergio.gonzalez.monroy@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
7 years agoapp/test: remove real PCI ids
Jan Viktorin [Tue, 14 Jun 2016 13:46:21 +0000 (15:46 +0200)]
app/test: remove real PCI ids

There are 2 new fake devices for testing PCI infra. All the fake devices
are now identified by non-existing vendor and device IDs so there is no
real driver to bind to them. The testing drivers match those IDs.

Suggested-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
7 years agoconfig: select maximum nodes and cores on ThunderX
Jerin Jacob [Mon, 13 Jun 2016 14:01:32 +0000 (19:31 +0530)]
config: select maximum nodes and cores on ThunderX

ThunderX platform can support dual sockets numa configuration and 48
cores per socket.

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
7 years agomk: define objcopy target and arch on IBM POWER
Chao Zhu [Wed, 15 Jun 2016 11:38:42 +0000 (19:38 +0800)]
mk: define objcopy target and arch on IBM POWER

This patch defines the target and arch value of objcopy program for
IBM POWER PPC64 little endian architecture.

Fixes: 99d6231fdca7 ("mk: define objcopy-specific target and arch")

Signed-off-by: Chao Zhu <chaozhu@linux.vnet.ibm.com>
7 years agoexamples/ip_pipeline: add core mappings script
Guruprasad Mukundarao [Fri, 10 Jun 2016 15:55:05 +0000 (16:55 +0100)]
examples/ip_pipeline: add core mappings script

This script parses the application configuration file and detects all the
pipelines specified therein, and then, it generates all the possible mappings
of those pipelines on the specified CPU core-list.

As a result, each of the possible pipeline-to-core mappings is saved as
separate output configuration file. For example- if input file is
edge_router_downstream.cfg with 3 pipeline (excluding pipeline 0) and
core-list is “1, 2”, following combinations will be generated-

Pipeline 1        Pipeline 2      Pipeline 3
Core = 1          Core = 1        Core = 2
Core = 1          Core = 2        Core = 1
Core = 2          Core = 1        Core = 1
Core = 2          Core = 2        Core = 1
Core = 2          Core = 1        Core = 2
Core = 1          Core = 2        Core = 2
Core = C1         Core = C1H      Core = C2
Core = C1         Core = C2       Core = C1H
Core = C2         Core = C1       Core = C1H

This script will help users to analyse the performance of application by
evaluating all the generated configuration files with different
pipelines-to-core mappings and obtaining the application configuration file
with best performance.

To run the script, issue the following command;

./pipeline-to-core-mapping.py -i <input_configuration_file> -pc "s<socket-id>c<core-id>"

Some optional arguments are as follows:
  -h, --help            show this help message and exit
  -ht {ON,OFF}, --hyper-thread {ON,OFF}
                        enable/disable hyper threading. default is ON
  -nO, --no-output-file
                        disable output config file generation. Output file
                        generation is enabled by default

Signed-off-by: Guruprasad Mukundarao <guruprasadx.rao@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
7 years agoexamples/ip_pipeline: fix false cacheline sharing among threads
Jasvinder Singh [Sun, 12 Jun 2016 12:42:47 +0000 (13:42 +0100)]
examples/ip_pipeline: fix false cacheline sharing among threads

In ip_pipeline app, the structure app_thread_data needs to be aligned to
the cache line boundary as threads on different cpu cores are accessing
fields of the app->thread_data and having this structure not aligned on
cacheline boundary leads to false cacheline sharing.

Fixes: 7f64b9c004aa ("examples/ip_pipeline: rework config file syntax")

Signed-off-by: Jasvinder Singh <jasvinder.singh@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
7 years agoexamples/ip_pipeline: fix null pointer dereference
Marcin Kerlin [Mon, 13 Jun 2016 09:35:48 +0000 (11:35 +0200)]
examples/ip_pipeline: fix null pointer dereference

Return value of function app_pipeline_type_find is not checking before
dereference. Fix this problem by adding checking condition.

Coverity issue: 127196
Fixes: b4aee0fb9c6d ("examples/ip_pipeline: reconfigure thread binding dynamically")

Signed-off-by: Marcin Kerlin <marcinx.kerlin@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
7 years agosched: fix build without red
Jasvinder Singh [Thu, 2 Jun 2016 10:28:10 +0000 (11:28 +0100)]
sched: fix build without red

This commit fixes the following compile error messages when
CONFIG_RTE_SCHED_RED=n and CONFIG_RTE_SCHED_COLLECT_STATS=y;

rte_sched.c: In function ‘rte_sched_port_update_subport_stats_on_drop’:
rte_sched.c:1090:41: error: unused parameter ‘red’
struct rte_mbuf *pkt, uint32_t red)
                                         ^
rte_sched.c: In function ‘rte_sched_port_update_queue_stats_on_drop’:
rte_sched.c:1116:39: error: unused parameter ‘red’
struct rte_mbuf *pkt, uint32_t red)

Fixes: 4d51afb5cdb6 ("sched: keep track of RED drops")

Signed-off-by: Jasvinder Singh <jasvinder.singh@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
7 years agokni: describe mempool capacity requirement
Alex Wang [Sat, 21 May 2016 07:58:36 +0000 (00:58 -0700)]
kni: describe mempool capacity requirement

Function like 'rte_kni_rx_burst()' keeps
allocating 'MAX_MBUF_BURST_NUM' mbufs to
kni fifo queue unless the queue's capacity
('KNI_FIFO_COUNT_MAX') is reached.  So, if
the mempool is under-provisioned, user may
run into "Out of Memory" logs from KNI code.
This commit documents the need to provision
mempool capacity of more than
"2 x KNI_FIFO_COUNT_MAX" for each KNI interface.

Signed-off-by: Alex Wang <alex@awakenetworks.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
7 years agokni: fix inverted function comments
Alex Wang [Sat, 21 May 2016 07:58:35 +0000 (00:58 -0700)]
kni: fix inverted function comments

The 'mbufs' alloc/free descriptions for
'rte_kni_tx_burst()' and 'rte_kni_rx_burst()'
should be inverted.

Fixes: 3fc5ca2 ("kni: initial import")

Signed-off-by: Alex Wang <alex@awakenetworks.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
7 years agodoc: add known issue in EAL argument parsing
Jingjing Wu [Mon, 23 May 2016 14:24:54 +0000 (22:24 +0800)]
doc: add known issue in EAL argument parsing

This patch docs the issue on EAL argument that the last EAL
argument is replaced by program name in argv[].

Reported-by: Ziye Yang <ziye.yang@intel.com>
Signed-off-by: Jingjing Wu <jingjing.wu@intel.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
7 years agopci: support class id probing
Ziye Yang [Tue, 24 May 2016 12:50:36 +0000 (20:50 +0800)]
pci: support class id probing

This patch is used to add the class_id (class_code,
subclass_code, programming_interface) support for
pci_device probe. With this patch, it will be
flexible for users to probe a class of devices
by class_id.

Signed-off-by: Ziye Yang <ziye.yang@intel.com>
7 years agoconfig: make libarchive optional
Thomas Monjalon [Tue, 14 Jun 2016 08:51:39 +0000 (10:51 +0200)]
config: make libarchive optional

The commit 66819e6 has introduced a dependency on libarchive to be able
to use some tar resources in the unit tests.
It is now an optional dependency because some systems do not have it
installed.

If CONFIG_RTE_APP_TEST_RESOURCE_TAR is disabled, the PCI test will not
be run. When a "configure" script will be integrated, the libarchive
availability could be checked to automatically enable the option.

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Reviewed-by: Jan Viktorin <viktorin@rehivetech.com>
7 years agoapp/test: fix wraparound of mempool performance rate
David Hunt [Thu, 26 May 2016 14:15:54 +0000 (15:15 +0100)]
app/test: fix wraparound of mempool performance rate

Recent CPU's can easily wrap around a 32-bit unsigned int in
the mempool perf test. Increase to a 64-bit uint.

Signed-off-by: David Hunt <david.hunt@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
7 years agoapp/test: fix bond device name too long
Michal Jastrzebski [Fri, 27 May 2016 15:20:52 +0000 (17:20 +0200)]
app/test: fix bond device name too long

Bond device name was too long (grather than 32 signs) that
cause mempool allocation to fail.

Fixes: 92073ef961ee ("bond: unit tests")

Signed-off-by: Michal Jastrzebski <michalx.k.jastrzebski@intel.com>
Acked-by: Bernard Iremonger <bernard.iremonger@intel.com>
Tested-by: Huilong Xu <huilongx.xu@intel.com>
7 years agoapp/test: fix array overflow warning with gcc 4.5
Tomasz Kulasek [Thu, 9 Jun 2016 11:14:32 +0000 (13:14 +0200)]
app/test: fix array overflow warning with gcc 4.5

app/test/test_cryptodev.c:
In function ‘create_snow3g_cipher_operation_oop.clone.15’:
x86_64-native-linuxapp-gcc/include/rte_memcpy.h:796:14
error: array subscript is above array bounds.

In test_cryptodev.c:
2429 rte_memcpy(sym_op->cipher.iv.data, iv, iv_len);

When iv_len is declared as 'unsigned int', rte_memcpy evaluates code for
buffer size bigger than 255, but while 'iv' array is 64 bytes long, it
causes 'above array bounds' warning in gcc 4.5 and breaks compilation.

Using uint8_t as a size of copied block prevents to evaluate in rte_memcpy
code for length bigger than 255, causing the problem.

The root of this issue and solution is the same as for commit 2c007ea10616
("app/test: fix array overflow warning with gcc 4.5")

Fixes: 9727af14b032 ("app/test: add out-of-place symmetric crypto operations")

Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
7 years agoapp/test: reduce duration of red functional test
Tomasz Kantecki [Mon, 13 Jun 2016 10:26:38 +0000 (11:26 +0100)]
app/test: reduce duration of red functional test

'red_autotest' changed to run only functional tests without test #4 which was
taking ~53 seconds. 'red_autotest' takes ~2[s] now.
'red_perf' has been added to run performance tests only).
'red_all' has been added to run all functional tests (including #4) and
perfromance tests. This reflects current 'red_autotest' behavior.

Other changes:
- machine TSC clock frequency detection takes place only once now.
- timeouts and number of iterations in functional tests have been reduced
  in order to shorten test duration.

Signed-off-by: Tomasz Kantecki <tomasz.kantecki@intel.com>
7 years agoapp/test: do not dump devices in PCI blacklist test
Jan Viktorin [Mon, 13 Jun 2016 15:07:45 +0000 (17:07 +0200)]
app/test: do not dump devices in PCI blacklist test

Dumping of devices in a unittest is useless. Instead, test whether
the test has been set up well - i.e. there are no devices.

Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
7 years agopci: allow to override sysfs path
Jan Viktorin [Mon, 13 Jun 2016 15:07:44 +0000 (17:07 +0200)]
pci: allow to override sysfs path

The SYSFS_PCI_DEVICES is a constant that makes the PCI testing
difficult as it points to an absolute path. We remove using this
constant and introducing a function pci_get_sysfs_path that gives
the same value. However, the user can pass a SYSFS_PCI_DEVICES env
variable to override the path. It is now possible to create a fake
sysfs hierarchy for testing.

Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
7 years agoapp/test: rename PCI blacklist test case
Jan Viktorin [Mon, 13 Jun 2016 15:07:43 +0000 (17:07 +0200)]
app/test: rename PCI blacklist test case

The current test_pci is just a single test case that tests the
blacklisting of devices. Rename it to test_pci_blacklist and call it
from the test_pci. The setup and cleanup are moved out of the
test_pci_blacklist entirely to cover all other tests.

Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
7 years agoapp/test: extract PCI setup and cleanup functions
Jan Viktorin [Mon, 13 Jun 2016 15:07:42 +0000 (17:07 +0200)]
app/test: extract PCI setup and cleanup functions

Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
7 years agoapp/test: use linked list to store PCI drivers
Jan Viktorin [Mon, 13 Jun 2016 15:07:41 +0000 (17:07 +0200)]
app/test: use linked list to store PCI drivers

The test unregisters all drivers before start. The drivers were stored
into a fixed-sized array. This is inflexible. This patch change this to
utilize a linked list for the same purpose.

Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
7 years agoapp/test: support resources archived by tar
Jan Viktorin [Mon, 13 Jun 2016 15:07:40 +0000 (17:07 +0200)]
app/test: support resources archived by tar

When a more complex resource (a file hierarchy) is needed, packing
every single file as a single resource would be very ineffective. For
that purpose, it is possible to pack the files into a tar archive,
extract it before test from the resource and finally clean up all the
created files.

This patch introduces functions resource_untar and resource_rm_by_tar
to perform those tasks. An example of using those functions is included
as a test.

A new dependency is required to build the app/test: libarchive.

Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
7 years agoapp/test: create files from resources
Jan Viktorin [Mon, 13 Jun 2016 15:07:39 +0000 (17:07 +0200)]
app/test: create files from resources

A resource can be written into the target filesystem by calling
resource_fwrite or resource_fwrite_file. Such file can be created
before a test is started and removed after the test finishes.

Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
7 years agoapp/test: support resources externally linked
Jan Viktorin [Mon, 13 Jun 2016 15:07:38 +0000 (17:07 +0200)]
app/test: support resources externally linked

To include resources from other source that the C source code we
can take advantage of the objcopy behaviour, i.e. packing of an
arbitrary file as an object file that is linked to the target program.

A linked object file is always accessible as a pair

extern const char beg_<name>;
extern const char end_<name>;
(extern const char siz_<name>;)

A unit test that packs the resource.c source file is included.

Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
7 years agoapp/test: introduce resources for tests
Jan Viktorin [Mon, 13 Jun 2016 15:07:36 +0000 (17:07 +0200)]
app/test: introduce resources for tests

Certain internal mechanisms of DPDK access different file system
structures (e.g. /sys/bus/pci/devices). It is difficult to test
those cases automatically by a unit test when such path is not
hard-coded and there is no simple way how to distribute fake ones
with the current testing environment.

This patch adds a possibility to declare a resource embedded in
the test binary itself. The structure resource cover the generic
situation - it provides a name for lookup and pointers to the
embedded data blob. A resource is registered in a constructor by
the macro REGISTER_RESOURCE.

Some initial tests of simple resources is included and added into
the group_1.

Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
7 years agomk: define objcopy-specific target and arch
Jan Viktorin [Mon, 13 Jun 2016 15:07:37 +0000 (17:07 +0200)]
mk: define objcopy-specific target and arch

The program objcopy uses non-standard conventions to name the
target and arch. Define the values for supported architectures
(tile and ppc_64 are missing).

Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
7 years agomk: fix missing internal dependencies
Christian Ehrhardt [Mon, 13 Jun 2016 11:06:57 +0000 (13:06 +0200)]
mk: fix missing internal dependencies

This patch adds missing DEPDIRS to avoid any library referring to
symbols they are not linked against.

Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
7 years agomk: fix vhost dependency to pthread
Christian Ehrhardt [Mon, 13 Jun 2016 11:06:56 +0000 (13:06 +0200)]
mk: fix vhost dependency to pthread

Add the missing external dependency to pthread to avoid referring to
symbols the library is not linked against.

Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
7 years agomk: reduce scope of whole-archive static linking
Ferruh Yigit [Fri, 10 Jun 2016 13:19:09 +0000 (15:19 +0200)]
mk: reduce scope of whole-archive static linking

The --whole-archive argument is only required for plugins (drivers)
and libraries used by these plugins.
Currently it covers all libraries.
Reducing the scope of this argument slightly reduce final application size
when statically linked.

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
7 years agomk: sort libraries in level order when linking
Thomas Monjalon [Fri, 10 Jun 2016 13:19:08 +0000 (15:19 +0200)]
mk: sort libraries in level order when linking

As stated in the comment:
    Order is important: from higher level to lower level

This is an attempt to make the layering order better respected.
It will help to restrict the --whole-archive scope for plugins.

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
7 years agomk: prevent overlinking in applications
Ferruh Yigit [Fri, 10 Jun 2016 13:19:07 +0000 (15:19 +0200)]
mk: prevent overlinking in applications

Replace --no-as-needed linker flag with --as-needed flag, which will
only link libraries directly called by application.
It can be achieved now that the libraries dependencies are handled
properly.

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
7 years agomk: remove library grouping during application linking
Thomas Monjalon [Fri, 10 Jun 2016 13:19:06 +0000 (15:19 +0200)]
mk: remove library grouping during application linking

>From "man ld":
  Using this option has a significant performance cost.
  It is best to use it only when there are unavoidable
  circular references between two or more archives.

Remove the option since it does not seem necessary.

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
7 years agomk: fix driver dependencies order for static application
Thomas Monjalon [Fri, 10 Jun 2016 13:19:05 +0000 (15:19 +0200)]
mk: fix driver dependencies order for static application

On a linker command line, the dependencies must be declared after
the libraries using them.
It will avoid some issues when building an application with static
libraries and --as-needed option.

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
7 years agomk: sort drivers in static application link list
Thomas Monjalon [Fri, 10 Jun 2016 13:19:04 +0000 (15:19 +0200)]
mk: sort drivers in static application link list

Just a clean up to prepare next patches.

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
7 years agomk: fix install with tar 1.29
Panu Matilainen [Mon, 23 May 2016 05:53:33 +0000 (08:53 +0300)]
mk: fix install with tar 1.29

--exclude became a positional option in tar 1.29, breaking the
test app filtering in "make install", causing .map files and all test
apps to get installed in bindir. Adjust the tar arguments accordingly,
this is compatible with older versions too since they do not care about
the order.

Ref: https://bugzilla.redhat.com/show_bug.cgi?id=1337864

Fixes: 6b62a72a70d0 ("mk: install a standard cutomizable tree")

Signed-off-by: Panu Matilainen <pmatilai@redhat.com>
7 years agoeal: remove useless includes of mempool and ring
Thomas Monjalon [Thu, 9 Jun 2016 14:40:01 +0000 (16:40 +0200)]
eal: remove useless includes of mempool and ring

The libraries rte_mempool and rte_ring are not used in EAL,
except for the ivshmem part (CONFIG_RTE_LIBRTE_IVSHMEM).

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: David Marchand <david.marchand@6wind.com>
7 years agolog: deprecate history dump
Thomas Monjalon [Thu, 9 Jun 2016 13:49:48 +0000 (15:49 +0200)]
log: deprecate history dump

The log history uses rte_mempool. In order to remove the mempool
dependency in EAL (and improve the build), this feature is deprecated.
The ABI is kept but the behaviour is now voided because it seems this
function was not used. The history can be read from syslog.

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: David Marchand <david.marchand@6wind.com>
7 years agomk: generate internal library dependencies
Panu Matilainen [Tue, 7 Jun 2016 10:01:54 +0000 (13:01 +0300)]
mk: generate internal library dependencies

Up to now dependencies between DPDK internal libraries have been
untracked at shared library level, requiring applications to know
about library internal dependencies and often consequently overlinking.

Since the dependencies are already recorded for build ordering in the
makefiles with DEPDIRS-y we can use that information to generate LDLIBS
entries for internal libraries automatically.

Also revert commit 8180554d82b3 ("vhost: fix linkage of driver with
library") which is made redundant by this change.

Signed-off-by: Panu Matilainen <pmatilai@redhat.com>
Acked-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
7 years agodoc: fix code section in versioning guide
John McNamara [Fri, 20 May 2016 14:08:39 +0000 (15:08 +0100)]
doc: fix code section in versioning guide

Fix broken console directive in the ABI validator section of the
ABI versioning docs.

Fixes: f1ef9794f9bd ("doc: add ABI guidelines")

Signed-off-by: John McNamara <john.mcnamara@intel.com>
7 years agodoc: fix testpmd commands for flow director
Mauricio Vasquez B [Tue, 7 Jun 2016 08:48:53 +0000 (10:48 +0200)]
doc: fix testpmd commands for flow director

A previous patch modified the CLIs witout updating the examples.

Fixes: 53b2bb9b7ea7 ("app/testpmd: new flow director commands")

Signed-off-by: Mauricio Vasquez B <mauricio.vasquezbernal@studenti.polito.it>
Acked-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
7 years agodoc: clarify port hotplug in testpmd
Pablo de Lara [Wed, 8 Jun 2016 15:40:17 +0000 (16:40 +0100)]
doc: clarify port hotplug in testpmd

Port needs to be stopped and then closed before it can be detached,
but the documentation was only saying to close the port.

Also, both sections for port detaching and attaching has been reformatted
slightly, to show clearly how to use the commands.

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
7 years agoapp/testpmd: fix typo when closing port
Pablo de Lara [Wed, 8 Jun 2016 15:20:31 +0000 (16:20 +0100)]
app/testpmd: fix typo when closing port

Fixes: ce8d561418d4 ("app/testpmd: add port configuration settings")

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
7 years agoapp/testpmd: prefetch data in the forwarding loop
Jerin Jacob [Tue, 3 May 2016 14:07:45 +0000 (19:37 +0530)]
app/testpmd: prefetch data in the forwarding loop

prefetch the next packet data address in advance in forwarding loop
for performance improvement.

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
7 years agoapp/testpmd: log mbuf pool creation
Olivier Matz [Mon, 30 May 2016 12:04:17 +0000 (14:04 +0200)]
app/testpmd: log mbuf pool creation

Enhance the logs related to mbuf pool creation. Display an info level
log when creating the mbuf, and display the error as a string on failure.

After the patch, we have:

  [...]
  EAL:   probe driver: 8086:10fb rte_ixgbe_pmd
  USER1: create a new mbuf pool <mbuf_pool_socket_0>: n=331456, \
      size=2176, socket=0
  EAL: Error - exiting with code: 1
    Cause: Creation of mbuf pool for socket 0 failed: Cannot allocate \
      memory

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
7 years agoapp/testpmd: add Tx flags in config command
Maciej Czekaj [Fri, 22 Apr 2016 14:51:20 +0000 (16:51 +0200)]
app/testpmd: add Tx flags in config command

"port config all txqflags <value>" allows for
specifying txq_flags value in command line.

Signed-off-by: Maciej Czekaj <maciej.czekaj@caviumnetworks.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
7 years agoapp/testpmd: add scatter enabling in config command
Maciej Czekaj [Fri, 22 Apr 2016 14:51:19 +0000 (16:51 +0200)]
app/testpmd: add scatter enabling in config command

"port config all scatter on|off" allows for
controlling rxmode.enable_scatter in command line.

Signed-off-by: Maciej Czekaj <maciej.czekaj@caviumnetworks.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
7 years agoapp/testpmd: add scatter enabling option
Maciej Czekaj [Fri, 22 Apr 2016 14:51:18 +0000 (16:51 +0200)]
app/testpmd: add scatter enabling option

This parameter allows for controlling rxmode.enable_scatter
which in turn allow for multi-segment packet receive tests.

Signed-off-by: Maciej Czekaj <maciej.czekaj@caviumnetworks.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
7 years agoexamples/quota_watermark: fix memory overflow
Piotr Azarewicz [Mon, 23 May 2016 12:19:56 +0000 (14:19 +0200)]
examples/quota_watermark: fix memory overflow

qw app at its init stage reserve 2*sizeof(int) memory space for quota
and low_watermark shared variables, but both apps (qw and qwctl) assign
wrong address for low_watermark pointer (out of reserved memzone space)
due to wrong pointer arithmetic.

CID 30709 : Extra sizeof expression (SIZEOF_MISMATCH)
suspicious_pointer_arithmetic: Adding 4UL /* sizeof (int) */ to pointer
(unsigned int *)(*qw_memzone).addr of type unsigned int * is suspicious
because adding an integral value to this pointer automatically scales
that value by the size, 4 bytes, of the pointed-to type, unsigned int.
Most likely, sizeof (int) is extraneous and should be replaced with 1.

Coverity issue: 30709
Fixes: 1d6c3ee3321a ("examples/quota_watermark: initial import")

Signed-off-by: Piotr Azarewicz <piotrx.t.azarewicz@intel.com>
7 years agoexamples/performance-thread: fix size of destination port ids
Tomasz Kulasek [Fri, 29 Apr 2016 16:30:24 +0000 (18:30 +0200)]
examples/performance-thread: fix size of destination port ids

After extending IPv4 next hop in lpm library, size of dst_port array was
changed from 16 to 32 bits in l3fwd-thread example, without modification
of the rest of path written for 16 bit value.

This patch uses similar approach for fix, like in commit 8353a36a9b4b
("examples/l3fwd: fix size of destination port ids"), restoring 16 bit size
for destination port ids and doing necessary conversion from 32 to 16 bit
after lpm_lookupx4.

Fixes: dc81ebbacaeb ("lpm: extend IPv4 next hop field")

Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
7 years agoexamples/performance-thread: fix destination port
Tomasz Kulasek [Fri, 29 Apr 2016 16:30:56 +0000 (18:30 +0200)]
examples/performance-thread: fix destination port

This fixes wrong logic in get_dst_port() on lpm path causing unpredictable
return value when ipv4/ipv6 lookup success (return with no value).

Fixes: dc81ebbacaeb ("lpm: extend IPv4 next hop field")

Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
7 years agoexamples/l3fwd: report error when no vector engine available
Jan Viktorin [Tue, 26 Apr 2016 11:30:59 +0000 (13:30 +0200)]
examples/l3fwd: report error when no vector engine available

If no SSE nor NEON are available the l3fwd should complain loudly
to quickly find out the reason.

Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
7 years agoexamples/l2fwd-jobstats: fix a typo
Rami Rosen [Thu, 21 Apr 2016 18:16:47 +0000 (21:16 +0300)]
examples/l2fwd-jobstats: fix a typo

Signed-off-by: Rami Rosen <rami.rosen@intel.com>
7 years agoexamples/l2fwd: increase mempool cache size for performance
Jerin Jacob [Tue, 19 Apr 2016 09:35:28 +0000 (15:05 +0530)]
examples/l2fwd: increase mempool cache size for performance

l3fwd sets the mempool cache size to 256, selected the same value for l2fwd

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Reshma Pattan <reshma.pattan@intel.com>
7 years agoexamples/l2fwd: remove hardcoded cycles per second
Jerin Jacob [Tue, 19 Apr 2016 09:35:27 +0000 (15:05 +0530)]
examples/l2fwd: remove hardcoded cycles per second

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Reshma Pattan <reshma.pattan@intel.com>
7 years agoexamples/ip_pipeline: add sample configs for various layers
Jasvinder Singh [Wed, 1 Jun 2016 14:01:02 +0000 (15:01 +0100)]
examples/ip_pipeline: add sample configs for various layers

The sample configuration file demonstrates that network layer components such
as TCP, UDP, ICMP etc, can be easily integrated into ip pipeline infrastructure.
Similarily, various other functionalities such as IP Reassembly for input
traffic with local destination and IP Fragmentation to enforce the MTU for
the routed output traffic, can be added using SWQs enabled with
reassembly and fragmentation features.

Signed-off-by: Jasvinder Singh <jasvinder.singh@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
7 years agoexamples/ip_pipeline: update routes when ports state change
Jasvinder Singh [Wed, 1 Jun 2016 14:01:01 +0000 (15:01 +0100)]
examples/ip_pipeline: update routes when ports state change

The routing pipeline registers a callback function with the nic ports and
this function is invoked for updating the routing entries (corresponding to
local host and directly attached network) tables whenever the nic ports
change their states (up/down).

Signed-off-by: Jasvinder Singh <jasvinder.singh@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
7 years agoexamples/ip_pipeline: assign MAC address to routing ports
Jasvinder Singh [Wed, 1 Jun 2016 14:01:00 +0000 (15:01 +0100)]
examples/ip_pipeline: assign MAC address to routing ports

As a result of tracking, output ports of routing pipelines are linked with
physical nic ports (potentially through other pipeline instances).
Thus, the mac addresses of the NIC ports are assigned to routing pipeline
out ports which are connected to them and are further used in routing table
entries instead of hardcoded default values.

Signed-off-by: Jasvinder Singh <jasvinder.singh@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
7 years agoexamples/ip_pipeline: link routing output ports to devices
Jasvinder Singh [Wed, 1 Jun 2016 14:00:59 +0000 (15:00 +0100)]
examples/ip_pipeline: link routing output ports to devices

This commit implements tracking mechanism for linking routing pipeline
output ports to physical NIC ports.

Once all the pipelines of the application are initialised, mechanism is
invoked during post initialisation phase for relating routing pipeline
output with NIC ports by navigating through the intermediate pipelines,
if present.

The tracking functions of the pipelines which help in navigating through
the intermediate pipelines are moved from pipeline_<pipeline_name>_be.c
to pipeline_<pipeline_name>.c. All pipelines except passthrough pipelines
use default tracking function (pipeline/pipeline_common_fe.c).

Signed-off-by: Jasvinder Singh <jasvinder.singh@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
7 years agoexamples/ip_pipeline: increase constants
Jasvinder Singh [Wed, 1 Jun 2016 14:00:58 +0000 (15:00 +0100)]
examples/ip_pipeline: increase constants

To allow more queues, pipeline types, threads, source/sink ports,etc., in
the ip pipeline application, larger values of macros are set.

Signed-off-by: Jasvinder Singh <jasvinder.singh@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>