1. Add fm10k_recv_scattered_pkts function to receive jumbo frame
and multi-segment packets.
2. Configure correct receive function in rx_init and dev_init.
Signed-off-by: Jeff Shaw <jeffrey.b.shaw@intel.com>
Signed-off-by: Chen Jing D(Mark) <jing.d.chen@intel.com>
1. Configure RSS in fm10k_dev_rx_init function.
2. Add fm10k_rss_hash_update and fm10k_rss_hash_conf_get to get
and inquery RSS configuration.
Signed-off-by: Jeff Shaw <jeffrey.b.shaw@intel.com>
Signed-off-by: Chen Jing D(Mark) <jing.d.chen@intel.com>
1. Add fm10k_recv_pkts and fm10k_xmit_pkts functions.
2. Link app function pointer to actual fm10k recv/xmit
functions.
3. Change Makefile to compile new file fm10k_rxtx.c
Signed-off-by: Jeff Shaw <jeffrey.b.shaw@intel.com>
Signed-off-by: Chen Jing D(Mark) <jing.d.chen@intel.com>
1. Add function to initialize RX queues.
2. Add function to initialize TX queues.
3. Add fm10k_dev_start, fm10k_dev_stop and fm10k_dev_close
functions.
4. Add function to close mailbox service.
Signed-off-by: Jeff Shaw <jeffrey.b.shaw@intel.com>
Signed-off-by: Chen Jing D(Mark) <jing.d.chen@intel.com>
1. Add 4 functions fm10k_dev_rx_queue_start,
fm10k_dev_rx_queue_stop, fm10k_dev_tx_queue_start,
and fm10k_dev_tx_queue_stop.
2. verify Rx packet buffer alignment is valid.
Hardware requires specific alignment for Rx packet buffers. At
least one of the following two conditions must be satisfied.
1) Address is 512B aligned
2) Address is 8B aligned and buffer does not cross 4K boundary.
Alignment is checked by the driver when the Rx queue is reset. It
is assumed that if an entire descriptor ring can be filled with
buffers containing valid alignment, then all buffers in that mempool
have valid address alignment. It is the responsibility of the user
to ensure all buffers have valid alignment, as it is the user who
creates the mempool.
It is assumed the buffer needs only to store a maximum size Ethernet
frame.
Signed-off-by: Jeff Shaw <jeffrey.b.shaw@intel.com>
Signed-off-by: Chen Jing D(Mark) <jing.d.chen@intel.com>
1. Add init function to scan and initialize fm10k PF device.
2. Add implementation to register fm10k pmd PF driver.
3. Add 3 functions fm10k_dev_configure, fm10k_stats_get and
fm10k_stats_get.
4. Add fm10k.h to define macros and basic data structure.
5. Add fm10k_logs.h to control log message output.
6. Change config/common_bsdapp and config/common_linuxapp, add
macros to control fm10k pmd driver compile for linux and bsd.
7. Add Makefile.
8. Change lib/Makefile to add fm10k driver into compile list.
9. Change mk/rte.app.mk to add fm10k lib into link.
10. Add ABI version of librte_pmd_fm10k
Signed-off-by: Jeff Shaw <jeffrey.b.shaw@intel.com>
Signed-off-by: Chen Jing D(Mark) <jing.d.chen@intel.com>
Signed-off-by: Michael Qiu <michael.qiu@intel.com>
Add fm10k device ID list into rte_pci_dev_ids.h.
Signed-off-by: Jeff Shaw <jeffrey.b.shaw@intel.com>
Signed-off-by: Chen Jing D(Mark) <jing.d.chen@intel.com>
Base driver is developed and maintained by Intel ND team, includes
basic functional service to Intel Ethernet Switch FM10000 Series
of silicons.
Any suggestion on bug fix and improvement within this directory is
welcome, but need this team to change and update.
Signed-off-by: Chen Jing D(Mark) <jing.d.chen@intel.com>
This could be useful to have this values for debug purposes.
Suggested-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
When offloading the checksums of ipip tunnels, m->l2_len is set to 0
as there is no tunnel or inner l2 header. Since this is a valid value
remove the test.
By the way, also remove the same test with l3_len because at this
point, it is expected that the software provides proper values in the
mbuf. It should avoid a test in dataplane processing and therefore
slightly increase performance.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
Advertise the DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM flag in the PMD
features. It means that the i40e PMD supports the offload of outer IP
checksum when transmitting tunneling packet.
Signed-off-by: Jijiang Liu <jijiang.liu@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
If the flag is advertised by a PMD, the NIC supports the outer IP
checksum TX offload of tunneling packets, therefore an application can
set the PKT_TX_OUTER_IP_CKSUM flag in mbufs when transmitting on this
port.
Signed-off-by: Jijiang Liu <jijiang.liu@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
Since previous commit, the flag PKT_TX_UDP_TUNNEL_PKT is not used by any PMD,
remove it from mbuf API and from csumonly (testpmd). In csumonly, the
PKT_TX_OUTER_IP_CKSUM flag is already set for vxlan checksum, providing
enough information to the underlying driver.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
The definition of the flag PKT_TX_UDP_TUNNEL_PKT in rte_mbuf.h was:
TX packet is an UDP tunneled packet. It must be specified when using
outer checksum offload (PKT_TX_OUTER_IP_CKSUM)
This flag was used to tell the NIC that the offload type is UDP
(I40E_TXD_CTX_UDP_TUNNELING flag). In the datasheet, it says it's
required to specify the tunnel type in the register. However, some tests
(see [1]) showed that it also works without this flag.
Moreover, it is not explained how the hardware use this
information. From a network perspective, this information is useless for
calculating the outer IP checksum as it does not depend on the payload.
Having this flag in the API would force the application to specify the
tunnel type for something that looks only useful for this PMD. It will
limit the number of possible tunnel types (we would need a flag for each
tunnel type) and therefore prevent to support outer IP checksum for
proprietary tunnels.
Finally, if a hardware advertises "I support outer IP checksum", it must
be supported for any payload types.
This has been validated by [2], knowing that the ipip test case was fixed
after this test report [3].
[1] http://dpdk.org/ml/archives/dev/2015-January/011380.html
[2] http://dpdk.org/ml/archives/dev/2015-January/011475.html
[3] http://dpdk.org/ml/archives/dev/2015-January/011610.html
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
From i40e datasheet:
The IP header type and its offload. In case of tunneling, the IIPT
relates to the inner IP header. See also EIPT field for the outer
(External) IP header offload.
00 - non IP packet or packet type is not defined by software
01 - IPv6 packet
10 - IPv4 packet with no IP checksum offload
11 - IPv4 packet with IP checksum offload
Therefore it is not needed to fill the IIPT field if no offload is
requested (we can keep the value to 00). For instance, the linux driver
code does not set it when (skb->ip_summed != CHECKSUM_PARTIAL). We can
do the same in the dpdk driver.
The function i40e_txd_enable_checksum() that fills the offload registers
can only be called for packets requiring an offload.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
The alias PKT_TX_IPV4_CSUM is only used in one place of i40e driver.
Remove it and only keep the legacy flag PKT_TX_IP_CSUM.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
In loopback mode, it's expected force link up even when there's no cable connect.
But in codes, setup_sfp() rewrites the related register.
It causes in the case 'multispeed_fiber', it can't link up without cable connect.
Signed-off-by: Cunming Liang <cunming.liang@intel.com>
Acked-by: Patrick Lu <patrick.lu@intel.com>
max_vfs will only be created by igb_uio driver, for other
drivers like vfio or pci_uio_generic, max_vfs will miss.
But sriov_numvfs is not driver related, just get the vf numbers
from that field.
Signed-off-by: Michael Qiu <michael.qiu@intel.com>
Acked-by: David Marchand <david.marchand@6wind.com>
cmdline_token_portlist_ops fell through cracks in the initial symbol
versioning patch, breaking pktgen build.
Signed-off-by: Panu Matilainen <pmatilai@redhat.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
To differentiate libraries that break ABI, we add a library version number
suffix to the library, which must be incremented when a given libraries ABI is
broken. This patch enforces that addition, sets the initial abi soname
extension to 1 for each library and creates a symlink to the base SONAME so that
the test applications will link properly.
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Sergio Gonzalez Monroy <sergio.gonzalez.monroy@intel.com>
Add linker version script files to each DPDK library to put a stake in the
ground from which we can start cleaning up API's
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Sergio Gonzalez Monroy <sergio.gonzalez.monroy@intel.com>
Add initial pass header files to support symbol versioning.
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Sergio Gonzalez Monroy <sergio.gonzalez.monroy@intel.com>
Hash filter control has been implemented for i40e. It includes
getting/setting,
- global hash configurations (hash function type, and symmetric
hash enable per flow type)
- symmetric hash enable per port
Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
In order to support hash filter configuration, filter type of hash
is added, also the corresponding structures, macros and definitions
are added.
Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Calculating the default RSS hash keys at run time is not needed
at all, and may have race conditions. The alternative is to use
array of random values which were generated manually as the
default hash keys.
Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
The Link bonding library is incorrectly using receive packet type flags
in the transmit policy hashing functions, which would cause packets
generated locally to be incorrectly distributed across the slave
devices. This patch completely removes the dependency on the packet
type flags and uses the ether_type from either the Ethernet header or
the VLAN headers for branching.
This patch also includes the associate changes in the test suite and in
the packet_burst_generator code to remove the dependences on the packet
type flags.
Signed-off-by: Declan Doherty <declan.doherty@intel.com>
Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
This is a duplication of some EAL parts for a standalone packaging
which is not documented.
Packaging should be done outside of DPDK.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
If at build phase we don't make any trie splitting,
then temporary build structures and resulting RT structure might be
much bigger than current.
>From other side - having just one trie instead of multiple can speedup
search quite significantly.
>From my measurements on rule-sets with ~10K rules:
RT table up to 8 times bigger, classify() up to 80% faster
than current implementation.
To make it possible for the user to decide about performance/space trade-off -
new parameter for build config structure (max_size) is introduced.
Setting it to the value greater than zero, instructs rte_acl_build() to:
- make sure that size of RT table wouldn't exceed given value.
- attempt to minimise number of tries in the table.
Setting it to zero maintains current behaviour.
That introduces a minor change in the public API, but I think the possible
performance gain is too big to ignore it.
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Vector code reorganisation/deduplication:
To avoid maintaining two nearly identical implementations of calc_addr()
(one for SSE, another for AVX2), replace it with a new macro that suits
both SSE and AVX2 code-paths.
Also remove no needed any more MM_* macros.
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Reorganise SSE code-path a bit by moving lo/hi dwords shuffle
out from calc_addr().
That allows to make calc_addr() for SSE and AVX2 practically identical
and opens opportunity for further code deduplication.
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Previous improvements made scalar method the fastest one
for tiny bunch of packets (< 4).
That allows us to remove specific vector code-path for small number of packets
(search_sse_2) and always use scalar method for such cases.
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Introduce new classify() method that uses AVX2 instructions.
>From my measurements:
On HSW boards when processing >= 16 packets per call,
AVX2 method outperforms it's SSE counterpart by 10-25%,
(depending on the ruleset).
When build with the compilers that don't support AVX2 instructions,
make rte_acl_classify_avx2() do nothing and return an error.
At runtime, if librte_acl was build with the compiler that supports AVX2,
this method is selected as default one on HW that supports AVX2.
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
New data type to manipulate 256 bit AVX values.
Rename field in the rte_xmm to keep common naming across SSE/AVX fields.
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Move common check for input parameters up into rte_acl_classify_alg().
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Make classify_scalar to behave in the same way as it's vector counterpart:
move match check out of the inner loop, etc.
That makes scalar and vector code look more identical.
Plus it improves scalar code performance.
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Right now we allocate indexes for all types of nodes, except MATCH,
at 'gen final RT table' stage.
For MATCH type nodes we are doing it at building temporary tree stage.
This is totally unnecessary and makes code more complex and error prone.
Rework the code and make MATCH indexes being allocated at the same stage
as all others.
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Introduced division of whole 256 child transition enties
into 4 sub-groups (64 kids per group).
So 2 groups within the same node with identical children,
can use one set of transition entries.
That allows to compact some DFA nodes and get space savings in the RT table,
without any negative performance impact.
>From what I've seen an average space savings: ~20%.
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
There was a bug at build phase that can cause matches beeing overwritten.
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Current rule-wildness based heuristics can cause unnecessary splits of
the ruleset.
That might have negative performance effect:
more tries to traverse, bigger RT tables.
After removing it, on some test-cases with big rulesets (~10K)
observed ~50% speedup.
No difference for smaller rulesets.
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Make data_indexes long enough to survive idle transitions.
That allows to simplify match processing code.
Also fix incorrect size calculations for data indexes.
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Fix compilation with RTE_LIBRTE_ACL_STANDALONE=y
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
For ixgbe, when queue start fails, for example, mbuf allocate
failure, the device will still start successfully, which could be
an issue.
Add return status check of queue start to avoid this issue.
Signed-off-by: Michael Qiu <michael.qiu@intel.com>
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
