Age | Commit message (Collapse) | Author | Files | Lines |
|
This patch adds dynamic arp support to UDP Replay VNF.
This patch fixes the performance issue observed with
arp support for UDP replay.
The arp support can be simply not used by not providing the
"-s option" which will revert to a plain UDP replay
functionality
Change-Id: I565eb83cd825470d4603ed5cb8fbe32f311eb0af
Signed-off-by: Vishwesh M Rudramuni <vishwesh.m.rudramuni@intel.com>
|
|
|
|
|
|
|
|
|
|
* changes:
vFW: ARPICMP packets shouldn't be processed by vFW
vFW: Remove TTL decrementing in firewall acting as gateway
|
|
JIRA: SAMPLEVNF-52
The stats counters are wrong when ARPICMP pkts are received by VFW
pipeline in case of HWLB mode as the vFW is trying to process
these pkts.
The ARPICMP pkts shouldn't be processed by VFW pipeline, as these
packets are processed by ARPICMP pipeline in case of HWLB.
Change-Id: Iefe947a9f63fd4943f37900582656a9a4f957f5d
Signed-off-by: Anand B Jyoti <anand.b.jyoti@intel.com>
|
|
JIRA: SAMPLEVNF-51
vFW is not functioning as a router and hence TTL decrement is
not required. Remove this TTL decrementing in vFW.
Change-Id: I49ae9b19a473eec0b8072f8a556dca299d89ef74
Signed-off-by: Anand B Jyoti <anand.b.jyoti@intel.com>
|
|
JIRA: SAMPLEVNF-47
The intermediate files and build directory will be regenerated
at build time. These files/folder are not required.
Change-Id: Iabf50c8dfcf80d9d6f4db0fca9a5a84e3fc6447a
Signed-off-by: Anand B Jyoti <anand.b.jyoti@intel.com>
|
|
|
|
|
|
|
|
|
|
|
|
JIRA: SAMPLEVNF-22
This patch support building the samplevnf with multiple different
versions of DPDK like 16.04, 16.11, 17.02 and 17.05.
Change-Id: Ic1a10442a03311010213c5054c8d459dd41c20cf
Signed-off-by: Anand B Jyoti <anand.b.jyoti@intel.com>
|
|
JIRA: SAMPLEVNF-20
The function name rte_hash_lookup_multi() is changed to
rte_hash_lookup_bulk() in DPDK version 16.11 and onwards.
This fix supports DPDK16.04, 16.11 and onwards and required for
DPDK17.05 migration.
Change-Id: I01f2145919768b5146116e67d7f349a5b039251c
Signed-off-by: Anand B Jyoti <anand.b.jyoti@intel.com>
|
|
JIRA: SAMPLEVNF-19
The rte_ring_sc_dequeue returns ENOENT in DPDK16.04 while ENOBUFS in
DPDK17.05. This leads to error in return value checking and mis behaviour.
Using of rte_ring_sc_dequeue() to be avoided to use the default
configuration as set during the creation of the queue as per the
DPDK API documentation. Similarly corrected for mp_dequeue as well.
Change-Id: Iacee1349b26d6ab432be891fad12313a6d68ca4d
Signed-off-by: Anand B Jyoti <anand.b.jyoti@intel.com>
|
|
JIRA: SAMPLEVNF-17
rte_eth_dev is removed in DPDK version 16.11 and onwards.
The corresponding error handling is fixed through samplevnf
gerrit 34667.
This patch is required for migration of VNFs to DPDK ver 17.05.
Change-Id: I088b705e3261227bcba7332aeb5a2a236e77a33e
Signed-off-by: Anand B Jyoti <anand.b.jyoti@intel.com>
|
|
JIRA: SAMPLEVNF-45
Apply the patches to disable the default DEBUG prints in the DPDK
libraries.
Change-Id: If17422e8398c8798ec9e97d9f141b2a8b2a53fd4
Signed-off-by: Anand B Jyoti <anand.b.jyoti@intel.com>
|
|
JIRA: SAMPLEVNF-45
The default RTE_LOG_LEVEL is set to 8 enables all debug prints
which impacts the performance testing.
This patch sets the RTE_LOG_LEVEL to INFO to avoid printing of
logs with level DEBUG.
Change-Id: I1fdeecab9cd82f37f62e1ccfa625f91d1cf16217
Signed-off-by: Anand B Jyoti <anand.b.jyoti@intel.com>
|
|
JIRA: SAMPLEVNF-46
Debug logs enabled by default causes delay in application initialization
as there can be many ACL rules to be setup during initialization.
This patch disables the ACL logs during rte_table_acl_build by default
and can be enabled by setting the RTE_LOG_LEVEL to DEBUG.
Change-Id: I48e642f8768135087b0ed4f1677c1c9ac9e0df58
Signed-off-by: Anand B Jyoti <anand.b.jyoti@intel.com>
|
|
JIRA: SAMPLEVNF-44
The UDP connection timer was getting reset for every packets with
reply traffic due to connection state check.
This patch avoids resetting the timer instead uses the expected
timeout to reset avoiding frequent timer reset for every packet.
Change-Id: I6b5aac62bb709f403aeacee75f52d099b2d2397f
Signed-off-by: Anand B Jyoti <anand.b.jyoti@intel.com>
|
|
JIRA: SAMPLEVNF-43
Before connection setup, if there are multiple packets in the burst
and are belongs to same connection, it was adding new connection for
the same 5 tuple connection, leading to multiple connections in the CT.
It was causing more number of active connections than possible.
Change-Id: I15c90e7feb465d96a9429ea6fedcb09fad9482e6
Signed-off-by: Anand B Jyoti <anand.b.jyoti@intel.com>
|
|
JIRA: SAMPLEVNF-41
Next generation platforms require more than 12 instances of VFW
to support more number of ports and throughout.
Increasing the max instances from 12 to 24 to support upto 200Gbps
in next generation platforms.
Change-Id: Ice4f84dfd4a96b84e7e5ef00349892d08e1d2348
Signed-off-by: Anand B Jyoti <anand.b.jyoti@intel.com>
|
|
|
|
JIRA: SAMPLEVNF-40
This patch adds sample vCGNAPT config for 1, 6, 10 port pair
Change-Id: I003e98b26e794d111af562232e8688f255a7d61a
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
|
|
JIRA: SAMPLEVNF-21
Removing unused function rte_pipeline_action_handler_port_ext() and
a patch rte_pipeline.patch which was applied on top of DPDK 16.04.
These are not used, hence cleaning up for DPDK migration.
Change-Id: I33a646ad27857c880969574b37412236ed3d780c
Signed-off-by: Anand B Jyoti <anand.b.jyoti@intel.com>
|
|
This patch enabled dynamic tune of timer timeout
[PIPELINE2]
type = TIMER
core = 2
timer_dyn_timeout = 20 #-- new config to change dyn timer timeout
n_flows = 1048576
Change-Id: I614cf73a7fc1afd9050580286c9a968aacb619fa
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
|
|
Change-Id: Iee26005d5e90a7481f6e6fb3c7441a2011fe5596
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
|
|
Use ETHER_TYPE instead packet_type when hw-ip-check is disabled.
Change-Id: Iced93b78d9f4e58011dd051e88ddd312112e0cba
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
|
|
JIRA: SAMPLEVNF-8
This code adds support for UDP Replay which acts as a reflector of all the
UDP packets on the requested port.
Change-Id: I633e65a4964fab4cc46e2ed09c742dcc63ce6258
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
|
|
This reverts commit 1b0802cbf10ac4d8911f0916d4ba0b02010924c8.
Change-Id: Ic2bb69125d0d849b260475026a1088911d2d7109
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
|
|
This reverts commit b04a6e5dacc458f81097c9b8a172dcbf16d8fd50.
Change-Id: Icb9a4ddea607c9dd9324a413407c266e8f6106f0
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
|
|
JIRA: SAMPLEVNF-10
LLC miss is possible when the connection data structure is accessed
after retrieved from hash table entries for TCP/UDP connection.
CPI for rte_ct_udp_packet() is not good due to if/else condition checks.
This patch provide a prefetch instructions to pre-fetch the cnxn data entry
structure to avoid the LLC miss and also optimize the if/else condition check
to improve the CPI for rte_ct_udp_packet() function.
Change-Id: Ieaf70815c919ddbe1ff34921222eac030fb41756
Signed-off-by: Anand B Jyoti <anand.b.jyoti@intel.com>
|
|
JIRA: SAMPLEVNF-18
This patch fixes segfault due to wrong use of portid while accessing
array
Change-Id: I1e8b57c27e7c26918bd61cb1d72dbe88ab269c98
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
|
|
|
|
This patch extends num_ports vnf can handle ports > 16 . This include
changes in the memory allocatiion and size of different paramters
Change-Id: Id8907327023ba863ad107bb874cfb60bd38daee5
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
|
|
|
|
|
|
|
|
Change-Id: I5eaf64a0cac8747ba5ac79bfd036a77cee766e39
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
|
|
JIRA: SAMPLEVNF-8
This code adds support for UDP Replay which acts as a reflector of all the
UDP packets on the requested port.
Change-Id: If8330580f58c785736e578e815d15d983ce20209
Signed-off-by: Vishwesh M Rudramuni <vishwesh.m.rudramuni@intel.com>
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
|
|
Change-Id: I9a3137f60b83c88b16027a9e6c404ccc3673b820
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
|
|
./tools/vnf_build.sh --silient
Change-Id: I50ecdc263a732d1ed823202107e49ffaa739e9b4
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
|
|
On SRIOV/OVS, NIC interface link always be UP. But to set the IP, we
need to bring down the interface. This patch allows the program to
continue after printing warning.
Change-Id: I79212f4d71501af5f3065143e067a22855a5fde6
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
|
|
JIRA: SAMPLEVNF-14
The simultaneous IPV4 and IPV6 traffic is not required/supported.
The IPv4 and IPv6 separation is done to avoid multiple checks between
v4 and v6 and to improve the throughput performance.
Remove this in-used/unsupported code from the acl library code.
Change-Id: I4b5c562e541b1616279f0332d8b837c44b7e865d
Signed-off-by: Anand B Jyoti <anand.b.jyoti@intel.com>
|
|
JIRA: SAMPLEVNF-14
The simultaneous IPV4 and IPV6 traffic is not required/supported.
The IPv4 and IPv6 separation is done to avoid multiple checks
between v4 and v6 and to improve the throughput performance.
This patch removes this unused/unsupported code from the vFW.
Change-Id: Iefea5a6ed8c9454be9807eb1796908aec7bfc199
Signed-off-by: Anand B Jyoti <anand.b.jyoti@intel.com>
|
|
JIRA: SAMPLEVNF-13
The packets drop counter of unsupported type was not cleared during
clearstats command of the vFW.
This patch adds the fix to clear the pkt_drop_unsupported_type
at the clearstats cmd.
Change-Id: I07d2c85a288efbc5e1fa6cee84c2ffe12e9673f4
Signed-off-by: Anand B Jyoti <anand.b.jyoti@intel.com>
|
|
JIRA: SAMPLEVNF-11
Enable Checksum
in makefile add CHECKSUM_REQ
--> hw_checksum_req=1 in config for hw_checksum
--> hw_checksum_req=0 in config for sw_checksum
disable checksum
in makefile comment CHECKSUM_REQ
Change-Id: I5ebcfe845082ea49acaa13bc963fc07b92b90650
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
|
|
JIRA: SAMPLEVNF-11
By Default CGNAPT will have only NAPT feature. To enable other advance
feature, Uncomment same in Makefile
- PCP
- SIP_ALG
- CT_CGNAT
- FTP_ALG
Change-Id: I8602726e6db8a899a94028f7bc16d6e7339438b9
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
|