Hemant Agrawal [Fri, 30 Jun 2017 08:54:17 +0000 (14:24 +0530)]
bus/fslmc: generic framework for mc object creation
There are muliple help mc object, which are not an independent
device, but they are required for dpaa2 based devices.
This framework allows registration and handling of all such
mc devices.
After test is finished the main thread waits only for lcores
where runners were started because, in case of using the
multicore scheduler, more cores are launched that do not need
to be waited for.
Signed-off-by: Kirill Rybalchenko <kirill.rybalchenko@intel.com> Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Pablo de Lara [Wed, 5 Jul 2017 05:26:18 +0000 (06:26 +0100)]
cryptodev: add mempool pointer in queue pair setup
The session mempool pointer is needed in each queue pair,
if session-less operations are being handled.
Therefore, the API is changed to accept this parameter,
as the session mempool is created outside the
device configuration function, similar to what ethdev
does with the rx queues.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Declan Doherty <declan.doherty@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com>
Change crypto device's session management to make it
device independent and simplify architecture when session
is intended to be used on more than one device.
Sessions private data is agnostic to underlying device
by adding an indirection in the sessions private data
using the crypto driver identifier.
A single session can contain indirections to multiple device types.
New function rte_cryptodev_sym_session_init has been created,
to initialize the driver private session data per driver to be
used on a same session, and rte_cryptodev_sym_session_clear
to clear this data before calling rte_cryptodev_sym_session_free.
Signed-off-by: Slawomir Mrozowicz <slawomirx.mrozowicz@intel.com> Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Declan Doherty <declan.doherty@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com>
Since crypto session will not be attached to a specific
device or driver, the field driver_id is not required
anymore (only used to check that a session was being
handled by the right device).
Signed-off-by: Slawomir Mrozowicz <slawomirx.mrozowicz@intel.com> Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Declan Doherty <declan.doherty@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com>
cryptodev: change attach session to queue pair API
Device id is going to be removed from session,
as the session will be device independent.
Therefore, the functions that attach/dettach a session
to a queue pair need to be updated, to accept the device id
as a parameter, apart from the queue pair id and the session.
Signed-off-by: Slawomir Mrozowicz <slawomirx.mrozowicz@intel.com> Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Declan Doherty <declan.doherty@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com>
cryptodev: do not create session mempool internally
Instead of creating the session mempool while configuring
the crypto device, apps will create the mempool themselves.
This way, it gives flexibility to the user to have a single
mempool for all devices (as long as the objects are big
enough to contain the biggest private session size) or
separate mempools for different drivers.
Also, since the mempool is now created outside the
device configuration function, now it needs to be passed
through this function, which will be eventually passed
when setting up the queue pairs, as ethernet devices do.
Signed-off-by: Slawomir Mrozowicz <slawomirx.mrozowicz@intel.com> Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Declan Doherty <declan.doherty@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com>
Cryptodev session structure was a duplication of the
cryptodev symmetric structure.
It was used by some PMDs that should use the symmetric
structure instead.
Since this structure was internal, there is no deprecation
notice required.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Declan Doherty <declan.doherty@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com>
Remove crypto device driver name string definitions from librte_cryptodev,
which avoid to library changes every time a new crypto driver was added.
The driver name is predefined internaly in the each PMD.
The applications could use the crypto device driver names based on
options with the driver name string provided in command line.
Changes device type identification to be based on a unique
driver id replacing the current device type enumeration, which needed
library changes every time a new crypto driver was added.
The driver id is assigned dynamically during driver registration using
the new macro RTE_PMD_REGISTER_CRYPTO_DRIVER which returns a unique
uint8_t identifier for that driver. New APIs are also introduced
to allow retrieval of the driver id using the driver name.
Signed-off-by: Slawomir Mrozowicz <slawomirx.mrozowicz@intel.com> Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Declan Doherty <declan.doherty@intel.com>
Pablo de Lara [Tue, 4 Jul 2017 00:12:43 +0000 (01:12 +0100)]
test/crypto: extend AES-GCM 192/256 to other PMDs
QAT, OpenSSL and DPAA2 PMDs support 192-bit and 256-bit key sizes
for AES-GCM, apart from 128-bit keys. Therefore, these tests that were
only used by AES-NI MB PMD are extended to these other two PMDs.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Sergio Gonzalez Monroy <sergio.gonzalez.monroy@intel.com>
Pablo de Lara [Tue, 4 Jul 2017 00:12:41 +0000 (01:12 +0100)]
test/crypto: rename some tests
AES-GCM tests are used for AES-NI MB, OpenSSL,
QAT and DPAA2_SEC PMDs, so they are not specific to the first one.
Therefore, _mb_ prefix can be removed.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Sergio Gonzalez Monroy <sergio.gonzalez.monroy@intel.com>
Pablo de Lara [Tue, 4 Jul 2017 00:12:40 +0000 (01:12 +0100)]
crypto/aesni_gcm: migrate to Multi-buffer library
Since Intel Multi Buffer library for IPSec has been updated to
support Scatter Gather List, the AESNI GCM PMD can link
to this library, instead of the ISA-L library.
This move eases the maintenance of the driver, as it will
use the same library as the AESNI MB PMD.
It also adds support for 192-bit keys.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Signed-off-by: Sergio Gonzalez Monroy <sergio.gonzalez.monroy@intel.com> Acked-by: Declan Doherty <declan.doherty@intel.com>
Pablo de Lara [Sun, 2 Jul 2017 02:55:05 +0000 (03:55 +0100)]
crypto/aesni_mb: support IPSec Multi-buffer lib v0.46
IPSec Multi-buffer library v0.46 has been released,
which includes, among othe features, support for 12-byte IV,
for AES-CTR, keeping also the previous 16-byte IV,
for backward compatibility reasons.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Declan Doherty <declan.doherty@intel.com>
Pablo de Lara [Sun, 2 Jul 2017 05:41:26 +0000 (06:41 +0100)]
cryptodev: use AES-GCM/CCM as AEAD algorithms
Now that all the structures/functions for AEAD algorithms
are in place, migrate the two supported algorithms
AES-GCM and AES-CCM to these, instead of using
cipher and authentication parameters.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Declan Doherty <declan.doherty@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com> Acked-by: Fiona Trahe <fiona.trahe@intel.com>
Pablo de Lara [Sun, 2 Jul 2017 05:41:21 +0000 (06:41 +0100)]
cryptodev: add AEAD parameters in crypto operation
AEAD operation parameters can be set in the new
aead structure, in the crypto operation.
This structure is within a union with the cipher
and authentication parameters, since operations can be:
- AEAD: using the aead structure
- Cipher only: using only the cipher structure
- Auth only: using only the authentication structure
- Cipher-then-auth/Auth-then-cipher: using both cipher
and authentication structures
Therefore, all three cannot be used at the same time.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Declan Doherty <declan.doherty@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com> Acked-by: Fiona Trahe <fiona.trahe@intel.com>
Pablo de Lara [Sun, 2 Jul 2017 05:41:19 +0000 (06:41 +0100)]
cryptodev: set AES-GMAC as auth-only algo
AES-GMAC is an authentication algorithm, based on AES-GCM
without encryption. To simplify its usage, now it can be used
setting the authentication parameters, without requiring
to concatenate a ciphering transform.
Therefore, it is not required to set AAD, but authentication
data length and offset, giving the user the option
to have Scatter-Gather List in the input buffer,
as long as the driver supports it.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Declan Doherty <declan.doherty@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com> Acked-by: Fiona Trahe <fiona.trahe@intel.com>
Pablo de Lara [Sun, 2 Jul 2017 05:41:16 +0000 (06:41 +0100)]
cryptodev: do not use AAD in wireless algorithms
For wireless algorithms (SNOW3G, KASUMI, ZUC),
the IV for the authentication algorithms (F9, UIA2 and EIA3)
was taken from the AAD parameter, as there was no IV parameter
in the authentication structure.
Now that IV is available for all algorithms, there is need
to keep doing this, so AAD is not used for these algorithms
anymore.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Declan Doherty <declan.doherty@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com> Acked-by: Fiona Trahe <fiona.trahe@intel.com>
Pablo de Lara [Sun, 2 Jul 2017 05:41:15 +0000 (06:41 +0100)]
cryptodev: add auth IV
Authentication algorithms, such as AES-GMAC or the wireless
algorithms (like SNOW3G) use IV, like cipher algorithms.
So far, AES-GMAC has used the IV from the cipher structure,
and the wireless algorithms have used the AAD field,
which is not technically correct.
Therefore, authentication IV parameters have been added,
so API is more correct. Like cipher IV, auth IV is expected
to be copied after the crypto operation.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Declan Doherty <declan.doherty@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com> Acked-by: Fiona Trahe <fiona.trahe@intel.com>
Pablo de Lara [Sun, 2 Jul 2017 05:41:14 +0000 (06:41 +0100)]
cryptodev: move IV parameters to session
Since IV parameters (offset and length) should not
change for operations in the same session, these parameters
are moved to the crypto transform structure, so they will
be stored in the sessions.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Declan Doherty <declan.doherty@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com> Acked-by: Fiona Trahe <fiona.trahe@intel.com>
Pablo de Lara [Sun, 2 Jul 2017 05:41:07 +0000 (06:41 +0100)]
cryptodev: add crypto op helper macros
In order to facilitate the access to the private data,
after the crypto operation, two new macros have been
implemented:
- rte_crypto_op_ctod_offset(c,t,o), which returns a pointer
to "o" bytes after the start of the crypto operation
(rte_crypto_op)
- rte_crypto_op_ctophys_offset(c, o), which returns
the physical address of the data "o" bytes after the
start of the crypto operation (rte_crypto_op)
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Declan Doherty <declan.doherty@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com> Acked-by: Fiona Trahe <fiona.trahe@intel.com>
Pablo de Lara [Sun, 2 Jul 2017 05:41:06 +0000 (06:41 +0100)]
cryptodev: remove useless alignment
rte_crypto_op and rte_crypto_sym_op structures
were marked as cache aligned.
However, since these structures are always initialized
in a mempool, this alignment is useless, since the mempool
forces the alignment of its objects.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Declan Doherty <declan.doherty@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com> Acked-by: Fiona Trahe <fiona.trahe@intel.com>
Pablo de Lara [Sun, 2 Jul 2017 05:41:05 +0000 (06:41 +0100)]
cryptodev: do not store pointer to op specific params
Instead of storing a pointer to operation specific parameters,
such as symmetric crypto parameters, use a zero-length array,
to mark that these parameters will be stored after the
generic crypto operation structure, which was already assumed
in the code, reducing the memory footprint of the crypto operation.
Besides, it is always expected to have rte_crypto_op
and rte_crypto_sym_op (the only operation specific parameters
structure right now) to be together, as they are initialized
as a single object in the crypto operation pool.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Declan Doherty <declan.doherty@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com> Acked-by: Fiona Trahe <fiona.trahe@intel.com>
Pablo de Lara [Sun, 2 Jul 2017 05:41:02 +0000 (06:41 +0100)]
cryptodev: move session type to generic crypto op
Session type (operation with or without session) is not
something specific to symmetric operations.
Therefore, the variable is moved to the generic crypto operation
structure.
Since this is an ABI change, the cryptodev library version
gets bumped.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Declan Doherty <declan.doherty@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com> Acked-by: Fiona Trahe <fiona.trahe@intel.com>
Allain Legacy [Fri, 23 Jun 2017 12:41:37 +0000 (08:41 -0400)]
net/virtio-user: fix crash when detaching device
The rte_eth_dev.data pointer is set to a reference to a static table.
Attempting to rte_free() it leads to a panic. For example, the
following commands result in a panic if run in testpmd
testpmd> port attach virtio_user0,path=/dev/vhost-net,iface=test0
testpmd> port stop 2
testpmd> port close 2
testpmd> port detach 2
Maxime Coquelin [Thu, 29 Jun 2017 07:58:10 +0000 (09:58 +0200)]
vhost: fix MTU device feature check
The MTU feature support check has to be done against MTU
feature bit mask, and not bit position.
Fixes: 72e8543093df ("vhost: add API to get MTU value") Cc: stable@dpdk.org Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com> Reviewed-by: Jianfeng Tan <jianfeng.tan@intel.com> Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Ivan Dyukov [Wed, 28 Jun 2017 12:40:31 +0000 (15:40 +0300)]
vhost: fix checking of device features
To compare enabled features in current device we must use bit
mask instead of bit position.
Fixes: c843af3aa13e ("vhost: access header only if offloading is supported") Cc: stable@dpdk.org Signed-off-by: Ivan Dyukov <i.dyukov@samsung.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com> Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Zhiyong Yang [Thu, 11 May 2017 02:16:35 +0000 (10:16 +0800)]
doc: update virtio ring size and header size
Add more explanations about vring size changes and different
virtio_header size.
Signed-off-by: Zhiyong Yang <zhiyong.yang@intel.com> Acked-by: John McNamara <john.mcnamara@intel.com> Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Jianfeng Tan [Wed, 7 Jun 2017 06:41:36 +0000 (06:41 +0000)]
vhost: fix TCP checksum
As PKT_TX_TCP_SEG flag in mbuf->ol_flags implies PKT_TX_TCP_CKSUM,
applications, e.g., testpmd, don't set PKT_TX_TCP_CKSUM when TSO
is set.
This leads to that packets get dropped in VM tcp stack layer because
of bad TCP csum.
To fix this, we make sure TCP NEEDS_CSUM info is set into virtio net
header when PKT_TX_TCP_SEG is set, so that VM tcp stack will not
check the TCP csum.
Fixes: 859b480d5afd ("vhost: add guest offload setting") Cc: stable@dpdk.org Signed-off-by: Jianfeng Tan <jianfeng.tan@intel.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com> Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Daniel Verkamp [Mon, 12 Jun 2017 21:29:04 +0000 (14:29 -0700)]
vhost: clean up per-socket mutex
vsocket->conn_mutex was allocated with pthread_mutex_init() but never
freed with pthread_mutex_destroy(). This is a potential memory leak,
depending on how pthread_mutex_t is implemented.
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com> Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
To make vectorized burst routines enabled, it is required to run on x86_64
architecture. If all the conditions are met, the vectorized burst functions
are enabled automatically. The decision is made individually on RX and TX.
There's no PMD option to make a selection.
When searching LKEY, if search key is mempool pointer, the 2nd cacheline
has to be accessed and it even requires to check whether a buffer is
indirect per every search. Instead, using address for search key can reduce
cycles taken. And caching the last hit entry is beneficial as well.
For Tx SW ring (txq->elts[]), indexes are kept and used in
txq->elts_head/tail. Because of this, one entry must always be left unused
and it also makes code complex. Changed to store counters instead of
indexes in order to make the code simpler and to reduce a few calculations.
Currently, if requested MTU is bigger than mbuf size and scattered
receive is not enabled, setting MTU to that value fails.
This patch allows setting this special MTU when device is stopped,
because scattered_rx will be re-configured during next port start
and driver may enable scattered receive according new MTU value.
After this patch, driver may select different receive function
automatically after MTU set, according MTU values selected.
Signed-off-by: Beilei Xing <beilei.xing@intel.com> Acked-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Tomasz Kulasek [Wed, 5 Jul 2017 18:54:29 +0000 (19:54 +0100)]
net/bonding: allow slaves to also be bonded devices
This patch removes restrictions in bonded device library which prevent a
bonded device to be added to another bonded device with the limitation
that 802.3ad mode is not supported if one or more slaves is also a
bonded device,
Signed-off-by: Declan Doherty <declan.doherty@intel.com> Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Tomasz Kulasek [Wed, 5 Jul 2017 18:54:28 +0000 (19:54 +0100)]
net/bonding: fix link properties management
This patch fixes the management of link properties in the bonded device.
In all mode except mode 4 a bonded device link will default to reporting
the link as full duplex and auto-neg. The link speed for a bond port is
calculated on it's active slaves and the particular mode it is running
in. The bonding link speed is reported based on the transmit link as in
some modes link speed between egress/ingress is not symmetrical.
- round-robin, balance, 802.3ad, TLB and ALB modes all report the link
speed as the sum of the speed of each active slave.
- active backup link speed is reported as the speed of the current
primary slave
- broadcast is reported as the minimum of value of the active slaves
link speeds.
In mode 4 (link aggregation 802.3ad) the properties of the first slave
added to the bonded device are slave and subsequent slaves are verified
to have the same properties.
Finally in the bond_ethdev_lsc_event_callback function the link
properties of the device are updated after any change to the number of
active slaves.
Fixes: 2efb58cbab6e ("bond: new link bonding library") Cc: stable@dpdk.org Signed-off-by: Declan Doherty <declan.doherty@intel.com> Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
This commit adds support for the flow API RSS action with the following
limitations:
- Only supported when isolated mode is enabled.
- The number of queues specified by the action (rte_flow_action_rss.num)
must be a power of two.
- Each queue index can be specified at most once in the configuration
(rte_flow_action_rss.queue[]).
- Because a queue can be associated with a single RSS context, it cannot
be targeted by multiple RSS actions simultaneously.
A special "parent" queue must be allocated in addition to a group of
standard Rx queues for RSS to work. This is done automatically outside of
isolated mode by the PMD when applications request several Rx queues.
Since each configured flow rule with the RSS action may target a different
set of queues, the PMD must have the ability to dynamically allocate
several parent queues, one per RSS group.
If isolated mode was requested the default RSS parent queue isn't created
in this case.
Refactor RSS parent queue allocations (currently limited to a single
parent) in preparation for flow API RSS action support.
Chas Williams [Thu, 29 Jun 2017 19:35:50 +0000 (15:35 -0400)]
net/vmxnet3: avoid code duplication
Refactor vmxnet3_post_rx_bufs() to call vmxnet3_renew_desc()
to update the newly allocated mbufs. While here, relocate the
relevant comments to vmxnet3_renew_desc().
Signed-off-by: Chas Williams <ciwillia@brocade.com> Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
In debug mode, all mbuf ol_flags are temporarily enabled while sitting
in the Rx queue to detect otherwise silent data corruption, however
some of them are special (indirect and control) and must be cleared
before returning mbufs to the pool to avoid crashing.
Tomasz Kulasek [Tue, 4 Jul 2017 16:46:27 +0000 (17:46 +0100)]
app/testpmd: add cmd for dedicated LACP Rx/Tx queues
Add new command to support enable/disable of dedicated Tx/Rx queue on
each slave of a bond device for LACP control plane traffic.
set bonding lacp dedicated_queues <port_id> [enable|disable]
When enabled this option creates dedicated queues on each slave device
for LACP control plane traffic. This removes the need to filter control
plane packets in the data path.
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com> Signed-off-by: Declan Doherty <declan.doherty@intel.com>
Tomasz Kulasek [Tue, 4 Jul 2017 16:46:26 +0000 (17:46 +0100)]
net/bonding: add dedicated HW queues for LACP control
Add support for hardware flow classification of LACP control plane
traffic to be redirect to a dedicated receive queue on each slave which
is not visible to application. Also enables a dedicate transmit queue
for LACP traffic which allows complete decoupling of control and data
paths.
This only applies to bonding devices running in mode 4
(link-aggregation-802.3ad).
Introduce two new APIs to support enable/disabled of dedicated
queues.
rte_eth_bond_8023ad_dedicated_queues_enable must be called before
bonding port is configured or started to reserved and configure the
dedicated queues.
When this option is enabled all slaves must support flow filtering
by Ethernet type and support one additional Tx and Rx queue on
each slave.
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com> Signed-off-by: Declan Doherty <declan.doherty@intel.com>
This patch fixes the maximum number of Tx an Rx queues supported by a
bonding device return by the rte_eth_dev_info_get function.
The bonding device now calculates the maximum number of supported Tx
and Rx queues based on the slaves bound to the bonded device, with the
minimum values of Tx and Rx queues from the device slaves being the
bonded devices maximum, as each slave must be able to support the same
number of Tx and Rx queues.
Fixes: 2efb58cbab6e ("bond: new link bonding library") Cc: stable@dpdk.org Signed-off-by: Declan Doherty <declan.doherty@intel.com>
git.droids-corp.org / dpdk.git / log