The implementation uses NEON gcc intrinsic.
Verified with testacl and acl_autotest applications on arm64 architecture.
Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
This implementation is based on IBM POWER version of
rte_cpuflags. We use software emulation of HW capability
registers, because those are usually not directly accessible
from userspace on ARM.
Signed-off-by: Vlastimil Kosar <kosar@rehivetech.com>
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
Acked-by: David Marchand <david.marchand@6wind.com>
Testpmd was only printing out second values when printing
RX/TX timestamp value, instead of both second and nanoseconds.
Since resolution of time counters is in nanoseconds,
testpmd should print out both.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Reviewed-by: John McNamara <john.mcnamara@intel.com>
The reorder library unit test was performed under the assumption that
the start sequence number was always 0.
This is not the case anymore as the start sequence number is initialized
by the first packet inserted into the reorder buffer.
This patch updates the unit test to reflect the new behavior.
Fixes: 7e1fa1de8a ("reorder: allow random number as starting point")
Reported-by: Mukesh Dua <mukesh.dua81@gmail.com>
Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
Acked-by: Sergio Gonzalez Monroy <sergio.gonzalez.monroy@intel.com>
One of the ways to reproduce the issue:
testpmd <EAL-OPTIONS> -- -i --txqflags=0
testpmd> set fwd txonly
testpmd> set txpkts 64,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4
testpmd> set txsplit rand
testpmd> start
After some time TX on ixgbe queue will hang,
and all packet transmission on that queue will stop.
This bug was first reported and investigated by
Vlad Zolotarov <vladz@cloudius-systems.com>:
"We can reproduce this issue when stressed the xmit path with a lot of highly
fragmented TCP frames (packets with up to 33 fragments with non-headers
fragments as small as 4 bytes) with all offload features enabled."
The root cause is that ixgbe_xmit_pkts() in some cases violates the HW rule
that the distance between TDs with RS bit set should not exceed 40 TDs.
>From the latest 82599 spec update:
"When WTHRESH is set to zero, the software device driver should set the RS bit
in the Tx descriptors with the EOP bit set and at least once in the 40
descriptors."
The fix is to make sure that the distance between TDs with RS bit set
would never exceed HW limit.
As part of that fix, tx_rs_thresh for ixgbe PMD is not allowed to be greater
then to 32 to comply with HW restrictions.
With that fix slight slowdown for the full-featured ixgbe TX path
might be observed (from our testing - up to 4%).
ixgbe simple TX path is unaffected by that patch.
Reported-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
For CSUM forwarding mode add ability to copy & split outgoing packet
into the new mbuf that consists of multiple segments.
For TXONLY and CSUM forwarding modes add ability to make number of
segments in the outgoing packet to vary on a per packet basis.
Number of segments and size of each segment is controlled by
'set txpkts' command.
Split policy is controlled by 'set txsplit' command.
Possible values are: on | off | rand.
Tha allows to increase test coverage for TX PMD codepaths.
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
It fixes compile issue on ICC 13.0.0.
Error logs:
app/test-pmd/cmdline.c(8160): error #188: enumerated type mixed
with another type
entry.input.flow.tunnel_flow.tunnel_type =
str2fdir_tunneltype(res->tunnel_type);
Fixes: 53b2bb9b7e ("app/testpmd: new flow director commands")
Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
icmp echo forward engine relies on the core rank rather than the number of
queues available.
This fix is still wrong, because if we set a nb-rxq > nb-txq, then we will
use an invalid txq, but with this, we are aligned on other fwd engines.
Signed-off-by: David Marchand <david.marchand@6wind.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
After implementing the fdir new modes for x550, the CLIs are modified.
Forgot to update the related help info and doc.
Fixes: 53b2bb9b7e ("app/testpmd: new flow director commands")
Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Hash_autotest is in test group 2 which had only 32 MB
of memory, which is increased to 64 MB to make it run.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
eal_flags_autotest was broken after commit
19bfa4dd ("eal: make the -n argument optional"),
since the unit test was checking that app would not run
if -n flag was missing, which now it is possible.
Also, subtest test_missing_n_flag() has been renamed
to test_invalid_n_flag(), as now -n flag is not compulsory.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
If using multiple cores on a system with hardware transactional
memory support, thread scaling does not work, as there was a single
point in the hash library which is a bottleneck for all threads,
which is the "free_slots" ring, which stores all the indices of
the free slots in the table.
This patch fixes the problem, by creating a local cache per logical core,
which stores locally indices of free slots,
so most times, writer threads will not interfere each other.
Fixes: 48a3991196 ("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>
Test commands are added to support selecting different length of GRE key.
Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Signed-off-by: Andrey Chilikin <andrey.chilikin@intel.com>
Test commands are added to support selecting input set, or adding
new input set per different pctype.
Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Signed-off-by: Andrey Chilikin <andrey.chilikin@intel.com>
Add a performance test for ring pmd, comparing performance of the pmd
compared to the basic rte_ring APIs.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
This patch adds a command to display DCB info in ports.
Signed-off-by: Jingjing Wu <jingjing.wu@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
Acked-by: Helin Zhang <helin.zhang@intel.com>
This patch changes the testpmd DCB forwarding stream to make it
based on traffic class.
It also fixes some coding style issues.
Signed-off-by: Jingjing Wu <jingjing.wu@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
Acked-by: Helin Zhang <helin.zhang@intel.com>
The "show port X rss-hash" command sometimes displays garbage instead of the
expected RSS hash key because the maximum key length is undefined. When the
requested key is too large to fit in the buffer,
rte_eth_dev_rss_hash_conf_get() does not update it.
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
DPDK uses a structure to get or set a new hash key (see
eth_rte_rss_hash_conf). rss_hf field from this structure is used in
rss_hash_get_conf to retrieve the hash key and in rss_hash_update uses
it to verify the key exists before trying to update it.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
For some ethnet-switch like intel RRC, all the packet forwarded
out by DPDK will be dropped in switch side, so the packet
generator will never receive the packet.
Signed-off-by: Michael Qiu <michael.qiu@intel.com>
Acked-by: Jijiang Liu <jijiang.liu@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
The different fdir mode needs different parameters, so, the parameter *mode*
is introduced to the CLI flow_director_filter and flow_director_mask. This
parameter can pormpt the user to input the appropriate parameters for different
mode.
Please be aware, as we should set the fdir mode, the value of the parameter
pkt-filter-mode, when we start testpmd. We cannot set a different mode for
mask or filter.
The new CLIs are added for the mac vlan and tunnel modes, like this,
flow_director_mask X mode MAC-VLAN vlan XXXX mac XX,
flow_director_mask X mode Tunnel vlan XXXX mac XX tunnel-type X tunnel-id XXXX,
flow_director_filter X mode MAC-VLAN add/del/update mac XX:XX:XX:XX:XX:XX
vlan XXXX flexbytes (X,X) fwd/drop queue X fd_id X,
flow_director_filter X mode Tunnel add/del/update mac XX:XX:XX:XX:XX:XX
vlan XXXX tunnel NVGRE/VxLAN tunnel-id XXXX flexbytes (X,X) fwd/drop queue X
fd_id X.
Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
There're fdir mask and supported flow type in the output of the CLI,
show port fdir. But not every parameter has meaning for all the fdir
modes, and the supported flow type is meaningless for mac vlan and
tunnel modes. So, we output different thing for different mode.
Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
For testpmd CLI's parameter pkt-filter-mode, there're new values supported for
fdir new modes, perfect-mac-vlan, perfect-tunnel.
Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
When a port is enabled, there're default values for the parameters of
fdir mask. For the new parameters, the default values also need to be
set.
Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
This is a useful default for simple applications where the assignment
of lcores to CPUs doesn't matter. It's also useful for more complex
applications that automatically assign tasks to cores based on the
NUMA topology.
Signed-off-by: Rich Lane <rich.lane@bigswitch.com>
Currently, there is a MAX_SOCKET macro which artificially limits the
number of NUMA sockets testpmd can use. Anything on a higher socket
ends up using socket zero. This patch replaces this with a variable
set during set_default_fwd_lcores_config() and uses RTE_MAX_NUMA_NODES
where a hard-coded max number of sockets is required.
Signed-off-by: Stephen Hurd <shurd@broadcom.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
'virtual' is a keyword and can't be used if the code is to compile with
C++ compilers.
If rte_devargs.h was included in C++ code, compilation with clang++
failed with an error. g++ did not fail, but only because of a bug
that treats it as an anonymous struct with a decl-specifier which it
ignores.
This simply renames the member to 'virt'.
Reported-by: Ming Zhao <mzhao@luminatewireless.com>
Signed-off-by: Christoph Gysin <christoph.gysin@gmail.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
Acked-by: David Marchand <david.marchand@6wind.com>
This patch relates to ABI change proposed for librte_port. Macros to
access the packet meta-data stored within the packet buffer has been
adjusted to cover the packet mbuf structure.
The LIBABIVER number is incremented.
Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
LPM table and pipeline apps have been modified to
include name parameter of the lpm table.
Signed-off-by: Jasvinder Singh <jasvinder.singh@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
The function rte_jhash2() was renamed rte_jhash_32b and
macros RTE_HASH_KEY_LENGTH_MAX and RTE_HASH_BUCKET_ENTRIES_MAX
were tagged as deprecated, so they can be removed in 2.2.
RTE_HASH_KEY_LENGTH is replaced in unit tests by an internal macro
for the memory allocation of all keys used.
The library version number is incremented.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
These functions were tagged as deprecated in 2.0 so they can be
removed in 2.2.
The library version is incremented.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Helin Zhang <helin.zhang@intel.com>
[Thomas: update doc and version]
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
The functions and structures are moved to app/test in order to keep
existing unit tests. Some minor changes were done in these functions
because of library scope restrictions.
An enum is also copied in two other applications to keep existing code.
The library version is incremented.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
The library version is incremented.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
The extended SCTP flow entries are now part of the standard API.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
The extended unified packet type is now part of the standard ABI.
As mbuf struct is changed, the mbuf library version is incremented.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
In table_autotest, the structures containing the parameters
to create the tables were not initialized, and therefore,
some checks could fail and so the unit test.
Also, due to this initialization, one of the tests that was
checking if the table was improperly initialized has been
removed, as it is not needed anymore (duplicated)
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
In commit: 1129992baa checking for offset alignment was removed.
Unit tests wasn't updated to reflect that change. This patch changes
checks with unaligned offsets to make tests pass.
Fixes: 1129992baa ("port: fix unaligned access to metadata")
Signed-off-by: Maciej Gajdzica <maciejx.t.gajdzica@intel.com>
Acked-by: Daniel Mrzyglod <danielx.t.mrzyglod@intel.com>
Test failed on verification if number of bytes
transmitted on each slave is not less than 90%
and greater than 110% of mean value of bytes transmitted
thru one slave. This was verified on a real system
but is difficult to achieve using virtualpmd.
That's why for unit tests only, it is sufficient to verify that with
high load (2 seconds transmission) all slaves are transmitting
so the traffic is balanced.
Fixes: 0c8396e6d7 ("bond: unit tests for mode 5")
Signed-off-by: Michal Jastrzebski <michalx.k.jastrzebski@intel.com>
'--no-huge' option now can workable with -m option.
Unit test for eal flag should change pass criterion.
Fixes: a7de7e6beb ("eal: allow combining -m and --no-huge")
Signed-off-by: Marvin Liu <yong.liu@intel.com>
In previous setting, mempool size and cache_size were both 32.
It does not satisfy with cache_size checking rule by now.
Cache size should be less than CONFIG_RTE_MEMPOOL_CACHE_MAX_SIZE and
mempool size / 1.5.
Fixes: 462321b44a ("mempool: limit cache size")
Signed-off-by: Marvin Liu <yong.liu@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
[Thomas: remove unused PKT_BURST_SZ]
Add new timer-manage race-condition test: We wrote a test to confirm
our suspicion that we could crash rte_timer_manage() under the right
circumstances. We repeatedly set several timers to expire at roughly
the same time on the master core. The master lcore just delays and runs
rte_timer_manage() about ten times per second. The slave lcores all
watch the first timer (timer-0) to see when rte_timer_manage() is
running on the master, i.e., timer-0's state is not PENDING.
At this point, each slave attempts to reset a subset of the timers to
a later expiration time. The goal here is to have the slaves moving
most of the timers to a different place in the master's pending-list,
while the master is traversing the same next-pointers (the slaves'
sl_next[0] pointers) and running callback functions. This eventually
results in the master traversing a corrupted linked-list.
In our observations, it results in an infinite loop.
Signed-off-by: Robert Sanford <rsanford@akamai.com>
Fix app/test timer stress test 2: Sometimes this test fails and
seg-faults because the slave lcores get out of phase with the master.
The master uses a single int, 'ready', to synchronize multiple slave
lcores through multiple phases of the test.
To resolve, we construct simple synchronization primitives that use one
atomic-int state variable per slave. The master tells the slaves when to
start, and then waits for all of them to finish. Each slave waits for
the master to tell it to start, and then tells the master when it has
finished.
Signed-off-by: Robert Sanford <rsanford@akamai.com>
