dpdk.git
8 years agoapp/testpmd: fix running without Xen whereas enabled
Christian Ehrhardt [Thu, 17 Mar 2016 14:47:03 +0000 (15:47 +0100)]
app/testpmd: fix running without Xen whereas enabled

With LIBRTE_PMD_XENVIRT enabled testpmd is built in a way to ONLY work
in XEN environments.
It will surface as:
   PMD: gntalloc: ioctl error
   EAL: Error - exiting with code: 1
     Cause: Creation of mbuf pool for socket 0 failed

With LIBRTE_PMD_XENVIRT enabled this now tries the xen style grant
table allocation, but falls back gracefully for the normal allocation.

The only thing left in the log will be the
   PMD: gntalloc: ioctl error

Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
8 years agovfio: fix resource leak
Daniel Mrzyglod [Mon, 11 Apr 2016 09:46:54 +0000 (11:46 +0200)]
vfio: fix resource leak

Coverity ID 13289: Resource leak:
The system resource will not be reclaimed and reused,
reducing the future availability of the resource.
In pci_vfio_get_group_fd: Leak of memory or pointers to system resources

Fixes: ff0b67d1c868 ("vfio: DMA mapping")

Signed-off-by: Daniel Mrzyglod <danielx.t.mrzyglod@intel.com>
Acked-by: Anatoly Burakov <anatoly.burakov@intel.com>
8 years agoexamples/vm_power_manager: fix libvirt dependency check
Marvin Liu [Mon, 11 Apr 2016 08:50:31 +0000 (16:50 +0800)]
examples/vm_power_manager: fix libvirt dependency check

vm_power_manager utilize libvirt API virDomainGetVcpuPinInfo to
retrieve domU vcpu information. This API is implemented from version 0.9.3.
Suse11 SP3 32bit default libvirt version is 0.8.8.

examples/vm_power_manager/channel_manager.c:
channel_manager.c:117:3: error: implicit declaration of function
'virDomainGetVcpuPinInfo'

Check and skip it from examples or raise an error when trying to compile
without libvirt or with a too old libvirt.

Fixes: e8ae9b662 ("examples/vm_power: channel manager and monitor in host")

Signed-off-by: Marvin Liu <yong.liu@intel.com>
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
8 years agoapp/test: fix memory corruption in reentrancy autotest
Olivier Matz [Mon, 11 Apr 2016 09:33:23 +0000 (11:33 +0200)]
app/test: fix memory corruption in reentrancy autotest

The mempool element size is set to 0, but 4 bytes are written in
my_obj_init():

  uint32_t *objnum = obj;
  memset(obj, 0, mp->elt_size);
  *objnum = i;

Change the MEMPOOL_ELT_SIZE constant to sizeof(uint32_t). This fixes
memory corruptions since we were writing outside of the object
boundaries.

Fixes: 104a92bd026 ("app: add reentrancy tests")

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
8 years agoexamples/vhost_xen: fix 32-bit build
Pablo de Lara [Sun, 10 Apr 2016 19:25:52 +0000 (20:25 +0100)]
examples/vhost_xen: fix 32-bit build

Compilation fails on 32 bits on Vhost_xen sample app, due to wrong casting:

examples/vhost_xen/vhost_monitor.c: In function ‘new_device’:
examples/vhost_xen/vhost_monitor.c:288:62: error: cast from pointer to integer
                               of different size [-Werror=pointer-to-int-cast]

Fixes: 47bd46112b71 ("xen: import xenvirt pmd and vhost_xen")

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
8 years agoxenvirt: fix 32-bit build
Pablo de Lara [Sun, 10 Apr 2016 19:25:51 +0000 (20:25 +0100)]
xenvirt: fix 32-bit build

Compilation fails on 32 bits on Xen driver, due to wrong casting:

drivers/net/xenvirt/virtqueue.h: In function ‘virtqueue_enqueue_xmit’:
drivers/net/xenvirt/virtqueue.h:234:24: error: cast from pointer to integer
                            of different size [-Werror=pointer-to-int-cast]
  start_dp[idx].addr  = rte_pktmbuf_mtod(cookie, uint64_t);
                        ^

Fixes: d6b324c00fc9 ("mbuf: get DMA address")

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
8 years agodoc: announce API changes for device objects
David Marchand [Thu, 7 Apr 2016 15:33:17 +0000 (17:33 +0200)]
doc: announce API changes for device objects

Following discussions with Jan, here is a deprecation notice to prepare for
hotplug and rte_device changes to come in 16.07.

Signed-off-by: David Marchand <david.marchand@6wind.com>
Acked-by: Jan Viktorin <viktorin@rehivetech.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
8 years agodoc: announce API changes for vhost refactoring
Yuanhan Liu [Wed, 6 Apr 2016 06:53:55 +0000 (14:53 +0800)]
doc: announce API changes for vhost refactoring

We currently exposed way too many fields (or even structures) than
necessary. For example, vhost_virtqueue struct should NOT be exposed
to user at all: application just need to tell the right queue id to
locate a specific queue, and that's all. Instead, the structure should
be defined in an internal header file. With that, we could do any changes
to it we want, without worrying about that we may offense the painful
ABI rules.

Similar changes could be done to virtio_net struct as well, just exposing
very few fields that are necessary and moving all others to an internal
structure.

Huawei then suggested a more radical yet much cleaner one: just exposing
a virtio_net handle to application, just like the way kernel exposes an
fd to user for locating a specific file, and exposing some new functions
to access those old fields, such as flags, virt_qp_nb.

With this change, we're likely to be free from ABI violations forever
(well, except when we have to extend the virtio_net_device_ops struct).
For example, following nice cleanup would not be a blocking one then:

    http://dpdk.org/ml/archives/dev/2016-February/033528.html

Suggested-by: Huawei Xie <huawei.xie@intel.com>
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Acked-by: Panu Matilainen <pmatilai@redhat.com>
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
8 years agodoc: announce ABI changes for user-owned mempool caches
Lazaros Koromilas [Tue, 5 Apr 2016 09:23:08 +0000 (12:23 +0300)]
doc: announce ABI changes for user-owned mempool caches

Deprecation notice for 16.04 for changes targeting release 16.07.
The changes affect struct rte_mempool, rte_mempool_cache and the
mempool API.

Signed-off-by: Lazaros Koromilas <l@nofutznetworks.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: David Hunt <david.hunt@intel.com>
8 years agomaintainers: claim responsibility for pcap PMD
Ferruh Yigit [Fri, 8 Apr 2016 17:03:28 +0000 (18:03 +0100)]
maintainers: claim responsibility for pcap PMD

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
8 years agodoc: fill nics features matrix for pcap
Ferruh Yigit [Fri, 8 Apr 2016 16:39:17 +0000 (17:39 +0100)]
doc: fill nics features matrix for pcap

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
8 years agodoc: add cryptodev chapter in prog guide
Declan Doherty [Fri, 8 Apr 2016 16:17:27 +0000 (17:17 +0100)]
doc: add cryptodev chapter in prog guide

Add a programmer's guide section for cryptodev library.

Signed-off-by: Declan Doherty <declan.doherty@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
8 years agodoc: add cryptodevs guide overview
Declan Doherty [Fri, 8 Apr 2016 16:17:26 +0000 (17:17 +0100)]
doc: add cryptodevs guide overview

Details supported device features and algorithms for each crypto PMD.

Signed-off-by: Declan Doherty <declan.doherty@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
8 years agoixgbe: fix packet type for VXLAN and NVGRE on X550
Wenzhuo Lu [Fri, 8 Apr 2016 08:10:24 +0000 (16:10 +0800)]
ixgbe: fix packet type for VXLAN and NVGRE on X550

VxLAN & NVGRE are supported by x550. As we know HW can parse
the packet and tell SW the type info. For VxLAN & NVGRE packets
there's some change. HW will not tell SW the info of the outer
header but the inner header instead. But we always take the
info as it's for the outer header. So the packet type info is
not right when x550 receives VxLAN & NVGRE packets.

As x550 only supports IPv4 VxLAN & NVGRE packets, we can tell
the outer header of VxLAN is IPv4 + UDP, and the outer header
of NVGRE is IPv4 only. What we don't know is if there's
optional field in the outer IPv4 header.

This patch implement the support of packet type for VxLAN &
NVGRE. And it fixes the wrong packet type issue either.

BTW:
It doesn't fix any existing commit as although it resolve an
issue it's more like a new feature but not a fix.

Reported-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
8 years agoexamples/performance-thread: restrict to x86_64
Pablo de Lara [Fri, 8 Apr 2016 16:20:15 +0000 (17:20 +0100)]
examples/performance-thread: restrict to x86_64

Performance-thread sample app is only supported for x86_64 targets,
so this commit adds a check to avoid compilation on other targets.

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
8 years agoexamples/ipsec-secgw: fix anonymous union initialization
Pablo de Lara [Fri, 8 Apr 2016 17:01:09 +0000 (18:01 +0100)]
examples/ipsec-secgw: fix anonymous union initialization

In icc 14.0, compilation was broken:

examples/ipsec-secgw/sa.c(212): error: a designator for an anonymous
union member can only appear within braces corresponding to that anonymous union
        .cipher = { RTE_CRYPTO_CIPHER_OP_ENCRYPT, RTE_CRYPTO_CIPHER_AES_CBC,
         ^

The member in anonymous union initialization should be inside '{}',
otherwise it will report an error.

Fixes: d299106e8e31 ("examples/ipsec-secgw: add IPsec sample application")

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
8 years agoexamples/ip_pipeline: fix 32-bit build
Jasvinder Singh [Fri, 8 Apr 2016 17:13:27 +0000 (18:13 +0100)]
examples/ip_pipeline: fix 32-bit build

error log:
ip_pipeline/pipeline/pipeline_routing_be.c:1537:
integer constant is too large for 'long' type

Fixes: 0ae7275810f1 ("examples/ip_pipeline: add more functions to routing pipeline")

Signed-off-by: Jasvinder Singh <jasvinder.singh@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
8 years agoexamples/ip_pipeline: fix size of allocated memory
Marcin Kerlin [Fri, 8 Apr 2016 13:32:24 +0000 (15:32 +0200)]
examples/ip_pipeline: fix size of allocated memory

CID 120150:
Wrong size of the allocated memory. Passing argument as size of pointer
(8UL) instead of size of structure app_pipeline_firewall_rule.

Fixes: 67ebdbef0c31 ("examples/ip_pipeline: add bulk update of firewall rules")

Signed-off-by: Marcin Kerlin <marcinx.kerlin@intel.com>
8 years agoversion: 16.04-rc4
Thomas Monjalon [Thu, 7 Apr 2016 21:42:07 +0000 (23:42 +0200)]
version: 16.04-rc4

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
8 years agodoc: announce API change for extended stats
Harry van Haaren [Tue, 5 Apr 2016 17:58:09 +0000 (18:58 +0100)]
doc: announce API change for extended stats

This patch adds a notice that the API for the xstats
functionality will be modified in the 16.07 release, with
no backwards compatibility planned as it would require
code duplication in each PMD that supports xstats.

Signed-off-by: Harry van Haaren <harry.van.haaren@intel.com>
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: Remy Horton <remy.horton@intel.com>
Acked-by: David Harton <dharton@cisco.com>
Acked-by: Maryam Tahhan <maryam.tahhan@intel.com>
8 years agodoc: announce ABI change for pcap in pipeline port
Fan Zhang [Thu, 31 Mar 2016 13:29:11 +0000 (14:29 +0100)]
doc: announce ABI change for pcap in pipeline port

Several new fields will be added to structure rte_port_source_params for
source port enhancement with pcap file reading support.

Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
Acked-by: Jasvinder Singh <jasvinder.singh@intel.com>
Acked-by: Piotr Azarewicz <piotrx.t.azarewicz@intel.com>
8 years agodoc: fix spellings
John McNamara [Thu, 7 Apr 2016 21:14:48 +0000 (22:14 +0100)]
doc: fix spellings

Fix some spelling errors and typos.

Signed-off-by: John McNamara <john.mcnamara@intel.com>
8 years agodoc: fill nics features matrix for bnx2x
Rasesh Mody [Thu, 7 Apr 2016 18:22:53 +0000 (11:22 -0700)]
doc: fill nics features matrix for bnx2x

Signed-off-by: Rasesh Mody <rasesh.mody@qlogic.com>
8 years agoscripts: test build with libsso
Thomas Monjalon [Thu, 10 Mar 2016 22:16:18 +0000 (23:16 +0100)]
scripts: test build with libsso

The SNOW3G PMD has a dependency on libsso which can be downloaded
after having passed some Intel validation about export restrictions.

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
8 years agoscripts: add verbose test build option
Thomas Monjalon [Mon, 21 Mar 2016 19:48:22 +0000 (20:48 +0100)]
scripts: add verbose test build option

The option -v enables the verbose mode when testing a build.

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
8 years agoscripts: hook build test config
Thomas Monjalon [Mon, 21 Mar 2016 22:14:58 +0000 (23:14 +0100)]
scripts: hook build test config

Insert a hook at the end of the config procedure, after having
adapted the configuration to the environment variables and the options
passed to the script.
It allows to better tune the automatic configuration of the build tests
in a function located in the devel config file.

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
8 years agoscripts: allow tuning any test build option
Thomas Monjalon [Tue, 29 Mar 2016 09:53:14 +0000 (11:53 +0200)]
scripts: allow tuning any test build option

Any build option can be enabled or disabled by appending the end
of its name to the config name.
Examples:
+INTRINSICS to enable CONFIG_RTE_FORCE_INTRINSICS
~RXTX_CALLBACKS to disable CONFIG_RTE_ETHDEV_RXTX_CALLBACKS

These builtin (lowercase) options are also added for convenience:
+debug to enable every debug options
+default to set target machine as default

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
8 years agoscripts: allow tuning build options per test target
Thomas Monjalon [Tue, 29 Mar 2016 09:04:32 +0000 (11:04 +0200)]
scripts: allow tuning build options per test target

The global variables are reloaded between each build to allow
having different config options based on DPDK_TARGET.

Some checks can now be removed from the script as they can
be done in the devel config file.

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
8 years agoscripts: stop build test after first error
Thomas Monjalon [Tue, 8 Mar 2016 15:43:50 +0000 (16:43 +0100)]
scripts: stop build test after first error

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
8 years agoscripts: test build of performance-thread example
Thomas Monjalon [Mon, 8 Feb 2016 13:41:00 +0000 (14:41 +0100)]
scripts: test build of performance-thread example

This example is not part of the baseline because of its
experimental state. That's why it must be tested separately.

Fixes: b700090c8cec ("examples/performance-thread: mark as experimental")

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
8 years agoscripts: remove legacy build method test
Thomas Monjalon [Tue, 29 Mar 2016 09:51:27 +0000 (11:51 +0200)]
scripts: remove legacy build method test

Building with "make install T=" is now deprecated.
The script will test only the standard "make" command.

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
8 years agoscripts: fix run in any directory
Thomas Monjalon [Mon, 14 Mar 2016 20:55:13 +0000 (21:55 +0100)]
scripts: fix run in any directory

The path to load-devel-config.sh must be relative to the script.

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
8 years agomk: show version as a decimal integer
Thomas Monjalon [Tue, 22 Sep 2015 14:49:23 +0000 (16:49 +0200)]
mk: show version as a decimal integer

In order to ease packaging support of changes in DPDK build system,
introduce a decimal integer to compare version numbers.
It does not show the minor numbers as it is not meaningful for packaging.

Usage for DPDK 16.04:
% make showversionum
1604

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
8 years agoexamples: fix draining all queues in l3fwd derivatives
Tomasz Kulasek [Thu, 7 Apr 2016 16:38:31 +0000 (18:38 +0200)]
examples: fix draining all queues in l3fwd derivatives

In l3fwd-acl and l3fwd-power not all tx ports was included in tx_port_id
array, used to periodically drain only available ports. This caused that
some packets can remain in buffer when application stops to receiving
packets or when size of burst is small.

Fixes: e2366e74e029 ("examples: use buffered Tx")

Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
8 years agoexamples/l2fwd-crypto: fix string overflow
Pablo de Lara [Thu, 7 Apr 2016 13:23:09 +0000 (14:23 +0100)]
examples/l2fwd-crypto: fix string overflow

When parsing crypto device type, the string was being copied
with strcpy(), which could overflow the destination buffer
(which is 32 byte long), so snprintf() should be used instead.

This fixes coverity issue 124575:
/examples/l2fwd-crypto/main.c: 1005 in l2fwd_crypto_parse_args_long_options()
>>>     CID 124575:    (STRING_OVERFLOW)
>>>     You might overrun the 32 byte fixed-size string
"options->string_auth_algo" by copying "optarg" without checking the length.
1005    strcpy(options->string_auth_algo, optarg);

Fixes: 49f79e86480d ("examples/l2fwd-crypto: add missing string initialization")

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Declan Doherty <declan.doherty@intel.com>
8 years agoapp/test: fix array overflow warning with gcc 4.5
Tomasz Kulasek [Thu, 7 Apr 2016 14:02:13 +0000 (16:02 +0200)]
app/test: fix array overflow warning with gcc 4.5

DPDK/app/test/test_cryptodev.c:
In function 'test_snow3g_encrypted_authentication.clone.3':
DPDK/x86_64-ivshmem-linuxapp-gcc/include/rte_memcpy.h:796:14: error:
array subscript is above array bounds

In lines like:
    rte_memcpy(sym_op->cipher.iv.data, iv, iv_len);
when "iv" is 64 bytes long array, and "iv_len" is "unsigned int",
compiler tries to evaluate also a code for array size larger than 255 bytes
long and reports error "array subscript is above array bounds" in line:
rte_memcpy.h:796
    rte_mov128((uint8_t *)dst + 128, (const uint8_t *)src + 128);
caused by evaluation to:
    rte_mov128((uint8_t *)sym_op->cipher.iv.data + 128, (const uint8_t *)iv
+ 128);
where "iv" is 64 bytes long buffer and "iv + 128" point out of it, gcc 4.5.

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

Fixes: 8bdf665fe6c0 ("app/test: add SNOW 3G")

Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
8 years agokni: fix possible deadlock
Ferruh Yigit [Thu, 7 Apr 2016 15:55:17 +0000 (16:55 +0100)]
kni: fix possible deadlock

netif_rx() should be used in interrupt context. Replace it with
netif_rx_ni() which is safe to use in process context.

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
8 years agovhost: enable guest notification only on enabled queues
Rich Lane [Thu, 7 Apr 2016 00:29:06 +0000 (17:29 -0700)]
vhost: enable guest notification only on enabled queues

If the vhost PMD were configured with more queues than the guest, the old
code would segfault in rte_vhost_enable_guest_notification due to a NULL
virtqueue pointer.

Fixes: ee584e9710b9 ("vhost: add driver on top of the library")

Signed-off-by: Rich Lane <rich.lane@bigswitch.com>
Tested-by: Ciara Loftus <ciara.loftus@intel.com>
Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
8 years agoivshmem: avoid infinite loop when concatenating segments
Anatoly Burakov [Thu, 7 Apr 2016 11:00:00 +0000 (12:00 +0100)]
ivshmem: avoid infinite loop when concatenating segments

This patch aligns the logic used to check for the presence of
adjacent segments in has_adjacent_segments() with the logic used
in cleanup_segments() when actually deciding to concatenate or
not a pair of segments. Additionally, adjacent segments are
no longer considered overlapping to avoid generating errors for
segments that can happily coexist together.

This fixes an infinite loop that happened when segments where
adjacent in their physical or virtual addresses but not in their
ioremap addresses: has_adjacent_segments() reported the presence
of adjacent segments while cleanup_segments() was not considering
them for concatenation, resulting in an infinite loop since the
result of has_adjacent_segments() is used in the decision to
continue looping in cleanup_segments().

Signed-off-by: David Verbeiren <david.verbeiren@intel.com>
Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
8 years agoethdev: fix unchecked return value
Slawomir Mrozowicz [Thu, 7 Apr 2016 11:46:32 +0000 (13:46 +0200)]
ethdev: fix unchecked return value

It fix coverity issue:
CID 124557 (#1 of 1): Unchecked return value (CHECKED_RETURN)
check_return: Calling rte_eth_tx_buffer_set_err_callback without
checking return value (as is done elsewhere 6 out of 7 times).

Fixes: d6c99e62c852 ("ethdev: add buffered Tx")

Signed-off-by: Slawomir Mrozowicz <slawomirx.mrozowicz@intel.com>
8 years agodoc: fix release notes for 16.04
John McNamara [Thu, 7 Apr 2016 16:02:33 +0000 (17:02 +0100)]
doc: fix release notes for 16.04

Fix grammar, spelling and formatting of DPDK 16.04 release notes.

Signed-off-by: John McNamara <john.mcnamara@intel.com>
8 years agodoc: add l2fwd-crypto sample app guide
Pablo de Lara [Thu, 7 Apr 2016 14:49:52 +0000 (15:49 +0100)]
doc: add l2fwd-crypto sample app guide

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Declan Doherty <declan.doherty@intel.com>
8 years agodoc: fix nics features matrix for mlx
Adrien Mazarguil [Thu, 7 Apr 2016 10:34:30 +0000 (12:34 +0200)]
doc: fix nics features matrix for mlx

Enable "other kdrv" because both NICs require kernel support to work.

Unicast and multicast MAC filters are also enabled as both address types can
be filtered on through the MAC add/remove/set callbacks.

Fixes: e86b85ca757b ("doc: fill nics features matrix for mlx")

Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
8 years agodoc: fix nics features matrix for ixgbe
Wenzhuo Lu [Thu, 7 Apr 2016 02:08:53 +0000 (10:08 +0800)]
doc: fix nics features matrix for ixgbe

Fixes: 83a4a15404ef ("doc: fill nics features matrix for e1000/igb and ixgbe")

Reported-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
8 years agodoc: fill nics features matrix for fm10k
Chen Jing D(Mark) [Thu, 7 Apr 2016 09:34:47 +0000 (17:34 +0800)]
doc: fill nics features matrix for fm10k

Add feature support list for fm10k, fm10k-vec, fm10kvf and
fm10kvf-vec.

Signed-off-by: Chen Jing D(Mark) <jing.d.chen@intel.com>
8 years agodoc: update multi process memory figure
Harry van Haaren [Wed, 6 Apr 2016 13:17:43 +0000 (14:17 +0100)]
doc: update multi process memory figure

This patch updates the titles in the multiprocess memory image
to read "Primary Process" and "Secondary Process" instead of
"DPDK Server Process" and "Customer Client Process".

The rest of the image has been converted from PNG to SVG.

Signed-off-by: Harry van Haaren <harry.van.haaren@intel.com>
8 years agodoc: add initialization section for crypto vdevs
Pablo de Lara [Mon, 4 Apr 2016 13:14:37 +0000 (14:14 +0100)]
doc: add initialization section for crypto vdevs

Explain how to create/initialize virtual crypto PMDs,
through command line and within an application.

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
8 years agodoc: fix typos in crypto titles
Pablo de Lara [Mon, 4 Apr 2016 13:14:35 +0000 (14:14 +0100)]
doc: fix typos in crypto titles

Fixes: eec136f3c54f ("aesni_gcm: add driver for AES-GCM crypto operations")
Fixes: 94b0ad8e0aa5 ("null_crypto: add driver for null crypto operations")

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
8 years agodoc: update libsso link for SNOW 3G
Pablo de Lara [Fri, 1 Apr 2016 11:23:57 +0000 (12:23 +0100)]
doc: update libsso link for SNOW 3G

A new process to request the libsso library required by the SNOW3G PMD
has been put in place, through a website, replacing the previous email method.
This commit updates the SNOW3G documentation, to reflect this change.

Since the library does not support newer gcc versions, the documentation
also contains a patch to make the library work with gcc > 5.0.

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Declan Doherty <declan.doherty@intel.com>
8 years agocryptodev: remove experimental label
Fiona Trahe [Wed, 6 Apr 2016 11:05:08 +0000 (12:05 +0100)]
cryptodev: remove experimental label

The cryptodev API was introduced in the DPDK 2.2 release.
Since then it has
 - been reviewed and iterated for the DPDK 16.04 release
 - had extensive use by the l2fwd-crypto app,
the ipsec-secgw example app,
the test app.
We believe it is now stable and the EXPERIMENTAL label should be removed.

Signed-off-by: Fiona Trahe <fiona.trahe@intel.com>
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
8 years agocryptodev: rename elements for clarity
Fiona Trahe [Mon, 4 Apr 2016 17:14:35 +0000 (18:14 +0100)]
cryptodev: rename elements for clarity

renamed rte_cryptodev_sym_session.type -> dev_type
(as it's not a session type, but a device type)

renamed rte_crypto_sym_op.type -> sess_type
(as it's not an op type, but a session type)

Signed-off-by: Fiona Trahe <fiona.trahe@intel.com>
8 years agoaesni_gcm: fix supported key sizes
Pablo de Lara [Wed, 6 Apr 2016 15:39:31 +0000 (16:39 +0100)]
aesni_gcm: fix supported key sizes

AES-GCM PMD only supports 128-bit keys, but not 192 and 256,
which the capabilities structure was reporting.

Fixes: 26c2e4ad5ad4 ("cryptodev: add capabilities discovery")

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
8 years agovhost: fix retrieval of numa node in driver
Ciara Loftus [Tue, 5 Apr 2016 16:09:47 +0000 (17:09 +0100)]
vhost: fix retrieval of numa node in driver

After some testing, it was found that retrieving numa information
about a vhost device via a call to get_mempolicy is more
accurate when performed during the new_device callback versus
the vring_state_changed callback, in particular upon initial boot
of the VM.  Performing this check during new_device is also
potentially more efficient as this callback is only triggered once
during device initialisation, compared with vring_state_changed
which may be called multiple times depending on the number of
queues assigned to the device.

Reorganise the code to perform this check and assign the correct
socket_id to the device during the new_device callback.

Fixes: ee584e9710b9 ("vhost: add driver on top of the library")

Signed-off-by: Ciara Loftus <ciara.loftus@intel.com>
Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
8 years agoapp/test: fix reentrancy autotest
Olivier Matz [Wed, 6 Apr 2016 13:28:01 +0000 (15:28 +0200)]
app/test: fix reentrancy autotest

The previous code in func_reentrancy autotest was doing in parallel
something close to:

  name = "common_name";
  do several times {
      obj = allocate_an_object(name)   // obj = ring, mempool, hash, lpm, ...
      if (obj == NULL && lookup(name) == NULL)
          return TEST_FAIL;
  }

This code is not safe. For instance:

   mempool_create() is called on core 0, it creates a ring. At the same
   time on core 1, mempool_create() is called too and the creation of the
   ring fails (EEXIST). But the mempool lookup can fail on core 1 if
   the mempool is not added in the list by core 0.

This commit fixes the func_reentrancy autotest that now works with all
tested class of objects.

Fixes: 104a92bd02 ("app: add reentrancy tests")

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
8 years agohash: fix race condition at creation
Olivier Matz [Wed, 6 Apr 2016 13:28:00 +0000 (15:28 +0200)]
hash: fix race condition at creation

To avoid a race condition while creating a new hash object, the
list has to be locked before the lookup, and released only once the
new object is added in the list.

As the lock is held by the rte_ring_create(), move its creation at the
beginning of the function and only take the lock after the ring is
created to avoid a deadlock.

Fixes: 48a3991196 ("hash: replace with cuckoo hash implementation")

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
8 years agohash: fix allocation of an existing object
Olivier Matz [Wed, 6 Apr 2016 13:27:59 +0000 (15:27 +0200)]
hash: fix allocation of an existing object

Change rte_hash*_create() functions to return NULL and set rte_errno to
EEXIST when the object name already exists. This is the behavior
described in the API documentation in the header file.

These functions were returning a pointer to the existing object in that
case, but it is a problem as the caller did not know if the object had
to be freed or not.

Doing this change also makes the hash API more consistent with the other
APIs (mempool, rings, ...).

Fixes: 916e4f4f4e ("memory: fix for multi process support")

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
8 years agolpm: fix allocation of an existing object
Olivier Matz [Wed, 6 Apr 2016 13:27:58 +0000 (15:27 +0200)]
lpm: fix allocation of an existing object

Change rte_lpm*_create() functions to return NULL and set rte_errno to
EEXIST when the object name already exists. This is the behavior
described in the API documentation in the header file.

These functions were returning a pointer to the existing object in that
case, but it is a problem as the caller did not know if the object had
to be freed or not.

Doing this change also makes the lpm API more consistent with the other
APIs (mempool, rings, ...).

Fixes: 916e4f4f4e ("memory: fix for multi process support")

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
8 years agoethdev: refine API to query supported packet types
Jianfeng Tan [Wed, 6 Apr 2016 03:51:13 +0000 (11:51 +0800)]
ethdev: refine API to query supported packet types

This change is to  make user code simpler. For PMDs which do not fill any
packet types, return 0 instead of -ENOTSUP as suggested by Bruce.

Usually, users only care if the required (by ptype_mask) ptypes can be
filled by the specified PMD. If the PMD implements dev_supported_ptypes_get
func is not important. And the introduce of another return value (-ENOTSUP)
would increase the complexity of user programs to check it.

Besides, there are ways to know if a PMD implements the func:
  a. see doc/guides/nics/overview.rst.
  b. use (~1) as parameter ptype_mask, then check if return 0.

Fixes: 78a38edf66de ("ethdev: query supported packet types")

Suggested-by: Bruce Richardson <bruce.richardson@intel.com>
Signed-off-by: Jianfeng Tan <jianfeng.tan@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
8 years agoena: fix build with icc
Ferruh Yigit [Wed, 6 Apr 2016 12:17:26 +0000 (13:17 +0100)]
ena: fix build with icc

With (ICC) 16.0.2 20160204, getting following warnings:

.../drivers/net/ena/base/ena_com.c(492): error #3656: variable
 "flags" may be used before its value is set
 ENA_SPINLOCK_LOCK(admin_queue->q_lock, flags);

.../drivers/net/ena/base/ena_com.c(1971): error #3656: variable
 "mem_handle" may be used before its value is set
 ENA_MEM_ALLOC_COHERENT(ena_dev->dmadev, len,

For both warnings the variable value is ignored, so there is no defect.
To comfort compiler warning, a initial value provided to variables.

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
8 years agoi40e: fix TSO
Zhe Tao [Wed, 6 Apr 2016 08:16:32 +0000 (16:16 +0800)]
i40e: fix TSO

Issue:

when using the following CLI in testpmd to enable ipv6 TSO feature
(set --txqflags=0 in the testpmd command)
set verbose 1
csum set ip hw 0
csum set udp hw 0
csum set tcp hw 0
csum set sctp hw 0
csum set outer-ip hw 0
csum parse_tunnel on 0
tso set 800 0
set fwd csum
start

We will not get what we want, the ipv6 packets sent out from IXIA can be
received by i40e, but cannot forward to another port.
The root cause is when HW doing the TSO offload for packets, it does not only
depends on the context descriptor to define the MSS and TSO payload size, it
also need to know whether this packets is ipv4 or ipv6, we use
i40e_txd_enable_checksum to generate the related fields for data descriptor.
But PMD will not call i40e_txd_enable_checksum if only the TSO offload flag is
set. The reason why ipv4 works fine for TSO in testpmd csum mode is csum engine
will set the ip csum flag when the packet is ipv4 and TSO is enabled but
will not set the flag for ipv6 and this flag will cause the
i40e_txd_enable_checksum to be invoked. For both the cases the TSO flag will be
set, so we need to use TSO flag to trigger the i40e_txd_enable_checksum.
The right logic here is we enable csum offload for both ipv4 and ipv6 when TSO
flag is set.

Fixes: e3f0151f ("i40e: enable Tx checksum only for offloaded packets")

Signed-off-by: Zhe Tao <zhe.tao@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
8 years agoi40evf: fix link info update
Jingjing Wu [Tue, 5 Apr 2016 02:01:34 +0000 (10:01 +0800)]
i40evf: fix link info update

The issue is the VF's link speed kept as 10G and status always was up.
It did not change even the physical link's status changed.
This patch fixes this issue to make VF's link info consistent with
physical link.

Fixes: 4861cde46116 ("i40e: new poll mode driver")

Signed-off-by: Jingjing Wu <jingjing.wu@intel.com>
8 years agoigb: fix i350 VF Rx
Wenzhuo Lu [Tue, 5 Apr 2016 01:10:00 +0000 (09:10 +0800)]
igb: fix i350 VF Rx

A problem is found on i350 VF. We found TX will happen once
per 4 packets. If only 1~3 packets are received, they will
not be forwarded. But the real problem is on RX side. The
reason is the default RX write-back threshold is changed to
4, so every first 3 packets may be hung there.

This patch checks the RX wthresh when setting up the RX
queue, and forces it to be 1, so every packet can be handled
immediately.

Fixes: 4a41c17dba18 ("igb: set default thresholds based on MAC type")

Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
8 years agovirtio: use zeroed memory for simple Tx header
Rich Lane [Tue, 5 Apr 2016 02:11:01 +0000 (19:11 -0700)]
virtio: use zeroed memory for simple Tx header

For simple TX the virtio-net header must be zeroed, but it was using memory
that had been initialized with indirect descriptor tables. This resulted in
"unsupported gso type" errors from librte_vhost.

We can use the same memory for every descriptor to save cachelines in the
vswitch.

Fixes: 6dc5de3a ("virtio: use indirect ring elements")

Signed-off-by: Rich Lane <rich.lane@bigswitch.com>
Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Acked-by: Jianfeng Tan <jianfeng.tan@intel.com>
8 years agovhost: fix error handling in destroy
Yuanhan Liu [Tue, 5 Apr 2016 06:00:43 +0000 (14:00 +0800)]
vhost: fix error handling in destroy

Fix following coverity defect:

    291     void
    292     vhost_destroy_device(struct vhost_device_ctx ctx)
    293     {
    294             struct virtio_net *dev = get_device(ctx);
    295
    >>>     CID 124565:  Null pointer dereferences  (NULL_RETURNS)
    >>>     Dereferencing a null pointer "dev".

Fixes: 45ca9c6f7bc6 ("vhost: get rid of linked list for devices")

Reported-by: John McNamara <john.mcnamara@intel.com>
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
8 years agoethdev: fix xstats retrieval with a null array
Olivier Matz [Mon, 4 Apr 2016 15:45:18 +0000 (17:45 +0200)]
ethdev: fix xstats retrieval with a null array

Coverity reports an issue in ethdev:

  *** CID 124562:  Null pointer dereferences  (FORWARD_NULL)
  /lib/librte_ether/rte_ethdev.c: 1518 in rte_eth_xstats_get()
  1512
  1513 /* global stats */
  1514      for (i = 0; i < RTE_NB_STATS; i++) {
  1515          stats_ptr = RTE_PTR_ADD(&eth_stats,
  1516
  rte_stats_strings[i].offset);
  1517 val = *stats_ptr;
  >>>     CID 124562:  Null pointer dereferences  (FORWARD_NULL)
  >>>     Dereferencing null pointer "xstats".
  1518                snprintf(xstats[count].name,
  sizeof(xstats[count].name),
  1519 "%s", rte_stats_strings[i].name);
  1520            xstats[count++].value = val;
  1521            }
  1522
  1523 /* per-rxq stats */

If a user calls rte_eth_xstats_get(portid, NULL, n) with n != 0,
it may result in a crash. Although the API documentation says that
n is the size of the table and xstats can be NULL if n == 0, we
can add an additional check here to make Coverity happy.

In that case, the return value is the same than when n == 0 is
passed, it returns the number of statistics.

Fixes: ce757f5c9a ("ethdev: new method to retrieve extended statistics")

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Harry van Haaren <harry.van.haaren@intel.com>
8 years agoapp/test: increase memory for group 2
Olivier Matz [Tue, 5 Apr 2016 07:36:41 +0000 (09:36 +0200)]
app/test: increase memory for group 2

The hash test (located in group_2) may require more than 64MB of memory,
especially if the memory is physically fragmented, making the test to
fail. So increase the memory to 128MB to avoid this issue.

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
8 years agoapp/test: fix EAL flags check
Olivier Matz [Mon, 4 Apr 2016 16:19:55 +0000 (18:19 +0200)]
app/test: fix EAL flags check

Since commit a88ba49e51, values larger than 4 are allowed,
the autotests need to be updated accordingly.

Fixes: a88ba49e51 ("config: fix CPU and memory parameters on IBM POWER8")

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
8 years agoexamples/l3fwd: fix crash with gcc 5
Tomasz Kulasek [Mon, 4 Apr 2016 14:45:23 +0000 (16:45 +0200)]
examples/l3fwd: fix crash with gcc 5

It seems that with gcc >5.x and -O2/-O3 optimization breaks packet
grouping algorithm.

When last packet pointer "lp" and "pnum->u64" buffer points the same
memory buffer, high optimization can cause unpredictable results.
It seems that assignment of precalculated group sizes may interfere
with initialization of new group size when lp points value inside
current group and didn't should be changed.

With gcc >5.x and optimization we cannot be sure which assignment will
be done first, so the group size can be counted incorrectly.

This patch eliminates intersection of assignment of initial group size
(lp[0] = 1) and precalculated group sizes when gptbl[v].idx < 4.

Fixes: 94c54b4158d5 ("examples/l3fwd: rework exact-match")

Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
8 years agodoc: fill nics features matrix for virtio
Jianfeng Tan [Tue, 5 Apr 2016 03:31:38 +0000 (11:31 +0800)]
doc: fill nics features matrix for virtio

Note: virtio is a para-virtualization device, which indicates that its
features depend on not only front end but also back end. Here by X, we
just mean the feature is supported in front end.

Signed-off-by: Jianfeng Tan <jianfeng.tan@intel.com>
8 years agodoc: fill nics features matrix for enic
John Daley [Tue, 5 Apr 2016 18:46:11 +0000 (11:46 -0700)]
doc: fill nics features matrix for enic

Signed-off-by: John Daley <johndale@cisco.com>
8 years agodoc: fill nics features matrix for e1000/igb and ixgbe
Wenzhuo Lu [Tue, 5 Apr 2016 02:54:10 +0000 (10:54 +0800)]
doc: fill nics features matrix for e1000/igb and ixgbe

Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
8 years agodoc: announce ABI change for PCI class
Ziye Yang [Wed, 17 Feb 2016 01:54:33 +0000 (09:54 +0800)]
doc: announce ABI change for PCI class

The purpose of this patch is used to add a new field
"class" in rte_pci_id structure. The new class field includes
class_id, subcalss_id, programming interface of a pci device.
With this field, we can identify pci device by its class info,
which can be more flexible instead of probing the device by
vendor_id OR device_id OR subvendor_id OR subdevice_id.
For example, we can probe all nvme devices by class field, which
can be quite convenient.

Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Helin Zhang <helin.zhang@intel.com>
Acked-by: Cunming Liang <cunming.liang@intel.com>
8 years agodoc: announce ABI changes for mempool allocation
Olivier Matz [Thu, 17 Mar 2016 09:05:11 +0000 (10:05 +0100)]
doc: announce ABI changes for mempool allocation

Add a deprecation notice for coming changes in mempool for 16.07.

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: David Hunt <david.hunt@intel.com>
Acked-by: Keith Wiles <keith.wiles@intel.com>
8 years agodoc: announce ABI changes for external mempool manager
David Hunt [Thu, 10 Mar 2016 11:55:40 +0000 (11:55 +0000)]
doc: announce ABI changes for external mempool manager

Announce the ABI breakage due to addition of external mempool
manager functionality which requires changes to rte_mempool
structure.

Signed-off-by: David Hunt <david.hunt@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Keith Wiles <keith.wiles@intel.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
8 years agodoc: announce ABI change for mempool cache
Keith Wiles [Fri, 12 Feb 2016 18:38:25 +0000 (12:38 -0600)]
doc: announce ABI change for mempool cache

Deprecation notice for 16.04 for changes to occur in
release 16.07 for rte_mempool memory reduction.

Signed-off-by: Keith Wiles <keith.wiles@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: David Hunt <david.hunt@intel.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
8 years agoexamples: fix build with icc 15.0.1
Daniel Mrzyglod [Mon, 4 Apr 2016 10:56:54 +0000 (12:56 +0200)]
examples: fix build with icc 15.0.1

error: loops in this subroutine are not good vectorization candidates
 (try compiling with O3 and/or IPO).

this error occurs in icc 15.0.1

Solution to disable this diagnostic message
https://software.intel.com/en-us/forums/intel-c-compiler/topic/537688

Fixes: d299106e8e31 ("examples/ipsec-secgw: add IPsec sample application")
Fixes: 8cc72f2814dd ("examples/vmdq_dcb: support X710")

Signed-off-by: Daniel Mrzyglod <danielx.t.mrzyglod@intel.com>
8 years agoexamples/l2fwd-crypto: fix build with icc
Daniel Mrzyglod [Mon, 4 Apr 2016 08:46:50 +0000 (10:46 +0200)]
examples/l2fwd-crypto: fix build with icc

Fix for compilation errors for icc:
error #188: enumerated type mixed with another type

Fixes: 00c58901f1b3 ("examples/l2fwd-crypto: use key-value list of supported algorithms")

Signed-off-by: Daniel Mrzyglod <danielx.t.mrzyglod@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
8 years agoversion: 16.04-rc3
Thomas Monjalon [Fri, 1 Apr 2016 21:52:02 +0000 (23:52 +0200)]
version: 16.04-rc3

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
8 years agodoc: fill nics features matrix for mlx
Adrien Mazarguil [Thu, 31 Mar 2016 10:22:37 +0000 (12:22 +0200)]
doc: fill nics features matrix for mlx

Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
8 years agomaintainers: claim responsibility for Intel i40e driver
Jingjing Wu [Tue, 29 Mar 2016 03:21:56 +0000 (11:21 +0800)]
maintainers: claim responsibility for Intel i40e driver

Signed-off-by: Jingjing Wu <jingjing.wu@intel.com>
Acked-by: Helin Zhang <helin.zhang@intel.com>
8 years agoexamples/ip_pipeline: fix pcap file parsing
Fan Zhang [Fri, 1 Apr 2016 09:55:30 +0000 (10:55 +0100)]
examples/ip_pipeline: fix pcap file parsing

This patch fixes the pcap file parsing in ip_pipeline. Originally, the
parser recognizes the pcap related entries regardless of the RTE_PORT_PCAP
macro definition status.

Fixes: fe5d04621303 ("examples/ip_pipeline: add pcap file dump")

Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
8 years agoport: clean up
Fan Zhang [Fri, 1 Apr 2016 13:41:57 +0000 (14:41 +0100)]
port: clean up

This patch clean-up the code in librte_port.
The clean-up includes the following:

* Clearer error message display.
* Remove unnecessary RTE_NEXT_ABI macro warping.
* Remove __rte_unused attribute

Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
8 years agoport: fix pcap sink parameter check
Fan Zhang [Fri, 1 Apr 2016 13:41:56 +0000 (14:41 +0100)]
port: fix pcap sink parameter check

This patch fixes sink port parameter checking logic.
Originally, if user set field "file_name" with meaning value
but leave PCAP support feature disabled, the program simply
ignores this field without notifying the user.

Fixes: eb5f4119b2bc ("port: add pcap file dump")

Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
8 years agoport: fix pcap source parameter check
Fan Zhang [Fri, 1 Apr 2016 13:41:55 +0000 (14:41 +0100)]
port: fix pcap source parameter check

This patch fixes source port parameter checking logic.
Originally, if user set field "file_name" with meaning value
but leave PCAP support feature disabled, the program simply
ignores this field without notifying the user.

Fixes: d4b42133d85b ("port: add pcap file source")

Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
8 years agoethdev: add 100G link speed
Thomas Monjalon [Thu, 31 Mar 2016 22:12:31 +0000 (00:12 +0200)]
ethdev: add 100G link speed

The link speed configuration is now done with bitmaps so 100G speed
requires only a new bit flag.
The actual link speed is a number so its size must be increased from
16-bit to 32-bit.

Signed-off-by: Marc Sune <marcdevel@gmail.com>
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Tested-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Tested-by: Matej Vido <vido@cesnet.cz>
8 years agoethdev: convert speed number to bitmap flag
Marc Sune [Thu, 31 Mar 2016 22:12:30 +0000 (00:12 +0200)]
ethdev: convert speed number to bitmap flag

It is a helper for the bitmap configuration.

Signed-off-by: Marc Sune <marcdevel@gmail.com>
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
8 years agoethdev: redesign link speed config
Marc Sune [Thu, 31 Mar 2016 22:12:29 +0000 (00:12 +0200)]
ethdev: redesign link speed config

This patch redesigns the API to set the link speed/s configuration
of an ethernet port. Specifically:

- it allows to define a set of advertised speeds for
  auto-negociation.
- it allows to disable link auto-negociation (single fixed speed).
- default: auto-negociate all supported speeds.

A flag autoneg in struct rte_eth_link indicates if link speed was a
result of auto-negociation or was fixed by configuration.

Signed-off-by: Marc Sune <marcdevel@gmail.com>
Tested-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Tested-by: Beilei Xing <beilei.xing@intel.com>
Tested-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
8 years agoethdev: add speed capabilities
Marc Sune [Thu, 31 Mar 2016 22:12:28 +0000 (00:12 +0200)]
ethdev: add speed capabilities

The speed capabilities of a device can be retrieved with
rte_eth_dev_info_get().

The new field speed_capa is initialized in the drivers without
taking care of device characteristics in this patch.
When the capabilities of a driver are accurate, the table in
overview.rst must be filled.

Signed-off-by: Marc Sune <marcdevel@gmail.com>
8 years agoethdev: rename link speed constants
Marc Sune [Thu, 31 Mar 2016 22:12:27 +0000 (00:12 +0200)]
ethdev: rename link speed constants

The speed numbers ETH_LINK_SPEED_ are renamed ETH_SPEED_NUM_.
The prefix ETH_LINK_SPEED_ is kept for AUTONEG and will be used
for bit flags in next patch.

Signed-off-by: Marc Sune <marcdevel@gmail.com>
8 years agoapp/testpmd: move speed and duplex parsing in a function
Marc Sune [Thu, 31 Mar 2016 22:12:26 +0000 (00:12 +0200)]
app/testpmd: move speed and duplex parsing in a function

The code for checking and parsing speed/duplex was duplicated.
The new function is also checking the speed/duplex combination.

Signed-off-by: Marc Sune <marcdevel@gmail.com>
8 years agoethdev: use constants for link duplex
Marc Sune [Thu, 31 Mar 2016 22:12:25 +0000 (00:12 +0200)]
ethdev: use constants for link duplex

Some duplex values are replaced from 0 to half-duplex when link is down.

Some drivers are still using their own constants for duplex modes.

Signed-off-by: Marc Sune <marcdevel@gmail.com>
8 years agoethdev: use constants for link state
Thomas Monjalon [Thu, 31 Mar 2016 22:12:24 +0000 (00:12 +0200)]
ethdev: use constants for link state

Define and use ETH_LINK_UP and ETH_LINK_DOWN where appropriate.

Signed-off-by: Marc Sune <marcdevel@gmail.com>
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
8 years agohash: fix multi-process support
Pablo de Lara [Fri, 1 Apr 2016 15:03:49 +0000 (16:03 +0100)]
hash: fix multi-process support

Hash library used a function pointer to choose a different
key compare function, depending on the key size.
As a result, multiple processes could not use the same hash table,
as the function addresses vary from one process to another.

Instead, a jump table is used, so each process has its own
function addresses, accessing this table with an index stored
in the hash table (note that using a custom key compare function
is not supported in multi-process mode).

Fixes: 48a399119619 ("hash: replace with cuckoo hash implementation")

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
8 years agohash: use common x86 flag
Pablo de Lara [Fri, 1 Apr 2016 15:03:48 +0000 (16:03 +0100)]
hash: use common x86 flag

Instead of using RTE_ARCH_X86_64, RTE_ARCH_X86_32
and RTE_ARCH_I686, use directly RTTE_ARCH_X86

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
8 years agoivshmem: fix race condition
Mauricio Vasquez B [Fri, 1 Apr 2016 11:22:12 +0000 (13:22 +0200)]
ivshmem: fix race condition

The memory zone could be freed just after adding it to the metadata
file and just before marking it as not freeable.
This patch changes the locking logic in order to prevent it.

Fixes: cd10c42eb5bc ("mem: fix ivshmem freeing")

Signed-off-by: Mauricio Vasquez B <mauricio.vasquezbernal@studenti.polito.it>
Acked-by: Anatoly Burakov <anatoly.burakov@intel.com>
8 years agobonding: fix loop boundary condition
Vladyslav Buslov [Fri, 1 Apr 2016 12:16:59 +0000 (15:16 +0300)]
bonding: fix loop boundary condition

Loop that calculates total number of tx descriptors in slave tx queues
should iterate up to nb_tx_queues, not nb_rx_queues.

Fixes: 3ef7955700e7 ("bonding: fix LACP mempool size")

Signed-off-by: Vladyslav Buslov <vladyslav.buslov@harmonicinc.com>
8 years agobonding: fix link detect in non-interrupt mode
Nelson Escobar [Sat, 26 Mar 2016 00:44:50 +0000 (17:44 -0700)]
bonding: fix link detect in non-interrupt mode

Stopping then re-starting a bond interface containing slaves that
used polling for link detection caused the bond to think all slave
links were down and inactive.

Move the start of the polling for link from slave_add() to
bond_ethdev_start() and in bond_ethdev_stop() make sure we clear
the last_link_status of the slaves.

Fixes: a45b288ef21a ("bond: support link status polling")

Signed-off-by: Nelson Escobar <neescoba@cisco.com>
Signed-off-by: John Daley <johndale@cisco.com>
Acked-by: Declan Doherty <declan.doherty@intel.com>
8 years agohash: fix typo in comment
Pablo de Lara [Tue, 29 Mar 2016 17:38:57 +0000 (18:38 +0100)]
hash: fix typo in comment

rte_hash_set_cmp_func() had an incorrect Doxygen comment
for one of its parameters.

Fixes: 95da2f8e9c61 ("hash: customize compare function")

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
8 years agoeal/ppc: fix prefetch instruction
Chao Zhu [Wed, 30 Mar 2016 15:39:17 +0000 (23:39 +0800)]
eal/ppc: fix prefetch instruction

Current prefetch instruction (dcbt) implementation for IBM POWER8 has wrong
Touch Hint(TH) parameter. The current setting of TH=1 indicates to load data from
current cache line and an unlimited number of sequentially following cache lines.
TTH=0 means to load data from current cache line. rte_prefetch0 function is defined
to load one cache line, which means TH=0 is suited here.

Signed-off-by: Chao Zhu <chaozhu@linux.vnet.ibm.com>
8 years agoconfig: fix CPU and memory parameters on IBM POWER8
Chao Zhu [Wed, 30 Mar 2016 15:39:16 +0000 (23:39 +0800)]
config: fix CPU and memory parameters on IBM POWER8

This patch fixes the max logic number and memory channel number settings
on IBM POWER8 platform.
1. The max number of logic cores of a POWER8 processor is 96. Normally,
   there are two sockets on a server. So the max number of logic cores
   are 192. So this parch set CONFIG_RTE_MAX_LCORE to 256.
2. The socket number on POWER8 little endian platform can be larger than 16.
   This patch set CONFIG_RTE_MAX_NUMA_NODES to 32 for POWER8.
3. Currently, the max number of memory channels are hardcoded to 4. However,
   on a POWER8 machine, the max number of memory channels are 8. This patch
   removes the constraint.

Signed-off-by: Chao Zhu <chaozhu@linux.vnet.ibm.com>