The log-level dpdk config value should be transparently
forwarded to DPDK via EAL argument. Since DPDK now supports
naming log-levels, VPP no longer needs to parse and call
rte_set_loglevel().
This was the other part of the DPDK log-level change.
It must have got missed during my initial checkin.
Without it passing dynamic log-level values like are silently
ignored.
Fixes: 6ca6ac6c88 ("dpdk: support passing log-level")
Change-Id: I732cec5f638c9924e3ffb04c4753f957e3633d64
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Upstream DPDK changed the names of the options to use dlopen()
with libibverbs and libmlx[45] from RTE_LIBRTE_MLX[45]_DLOPEN_DEPS
to RTE_IBVERBS_LINK_DLOPEN (handles both mlx4 and mlx5).
VPP's build option to enable this configuration when building DPDK
no longer worked starting when VPP moved to DPDK 19.02. Update VPP's
build options to enable the correct option name.
Change-Id: I8e34e1d3fc4ee8aac4fd6f2a7d27177f2b0dea50
Signed-off-by: Matthew Smith <mgsmith@netgate.com>
Moved code to the ethernet input node, and the interface output
path(s). Since we no longer skip ethernet-input, there's no reason
for device drivers to know anything about pcap rx tracing, etc.
Change-Id: I08d32fb1b90cbee1bd4f609837d533e047b36fa4
Signed-off-by: Dave Barach <dave@barachs.net>
Recent VIC models can support 25, 50, and 100Gbps links. Use the
helper (port_type_from_link_speed) to set the port type as it supports
all possible link speeds.
Change-Id: I748d8ac716a6393d116a9db8a599151c70a9000a
Signed-off-by: Hyong Youb Kim <hyonkim@cisco.com>
1) stats are accessed via the stat segment which is more condusive to
monitoring
2) stats are accurate in the presence of multiple threads. There's no
guarantee that an SA is access from only one worker.
Change-Id: Id5e217ea253ddfc9480aaedb0d008dea031b1148
Signed-off-by: Neale Ranns <nranns@cisco.com>
don't have to convert from mbuf to vlib_buffer then buffer index
save a few clock cycles in crypto-input
plus, a bit improvements of CLI
1. show more information, resource placement & qp stats
2. clear dpdk qp statistics
cleanup cli as sugguested by Sergio Gonzalez Monroy
Change-Id: Ic4fd65bfa9a6b05b344a9a40c554990dde072d19
Signed-off-by: Kingwel Xie <kingwel.xie@ericsson.com>
in the same maaner as with other tunnel tyeps we use
the FIB to cache and track the destination used to reach
the tunnel endpoint. Post encap we can then ship the packet
straight to this adjacency and thus elide the costly second
lookup.
- SA add and del function so they can be used both directly
from the API and for tunnels.
- API change for the SA dump to use the SA type
- ipsec_key_t type for convenience (copying, [un]formating)
- no matching tunnel counters in ipsec-if-input
Change-Id: I9d144a59667f7bf96442f4ca66bef5c1d3c7f1ea
Signed-off-by: Neale Ranns <nranns@cisco.com>
DPDK 19.02 adds two new fields to struct rte_cryptodev_qp_conf,
which the current code was not initializing properly.
Also session mempools are now required to have specific private data.
For that just use the new API to create symmetric session pools.
Change-Id: Ie732d4e10b908aeaea322717d6011113e3e7172c
Signed-off-by: Sergio Gonzalez Monroy <sgmonroy@gmail.com>
dpdk_crypto_input_trace was called before vlib_buffer_enqueue_to_next
then VLIB_FRAME_TRACE of next_frame->flag will be overwritten by
vlib_next_frame_change_ownership(), leading to a broken trace.
now it is working:
Packet 1
00:00:15:654983: dpdk-crypto-input
dev_id 0 next-index 1
00:00:15:654999: ip4-lookup
fib 0 dpo-idx 0 flow hash: 0x00000000
IPSEC_ESP: 18.1.0.71 -> 18.1.0.241
tos 0x00, ttl 254, length 168, checksum 0x96ea
......
Change-Id: I73d77c06c11db8911866adb6240b2565b690f469
Signed-off-by: Kingwel Xie <kingwel.xie@ericsson.com>
HQoS requires fixes to work with dpdk 19.02 so code is disabled and
pending deprecation unless active maintainer is found.
Change-Id: I3569c4287b6dfdd2c29e02375eb53bf01fa6ae84
Signed-off-by: Damjan Marion <damarion@cisco.com>
- DPDK overwrites metadata as part of rte_pktmbuf_init(...) so we need
reinitialize it
- additional checks added to ensure ref_count is never < 1
Change-Id: Ida336f81c4723e8f2e0ad4a70cb7b1ecfff978a0
Signed-off-by: Damjan Marion <damarion@cisco.com>
This patch introduces following changes:
- deprecated free lists which are not used and not compatible
with external buffer managers (i.e. DPDK)
- introduces native support for per-numa buffer pools
- significantly improves performance of buffer alloc and free
Change-Id: I4a8e723ae47056717afd6cac0efe87cb731b5be7
Signed-off-by: Damjan Marion <damarion@cisco.com>
Previous scheme was not cross-compile friendly...
Change-Id: Ib103b136231673157a02b8750312aa6073052c7e
Signed-off-by: Damjan Marion <damarion@cisco.com>
as this node is refactored in new style with
vlib_buffer_enqueue_to_next, we have to check if the 'count'
is greater than 0. otherise, the next_index would be invalid
then lead to a crash
Change-Id: If7c323b59c02b5c16bd9d77b65c946512cc972c1
Signed-off-by: Kingwel Xie <kingwel.xie@ericsson.com>
Storing buffer in local template seems to be better option....
Change-Id: I1a2fdd68cb956f99a5b36d2cd810fc623e089bcf
Signed-off-by: Damjan Marion <damarion@cisco.com>
Some dpdk driver does not support rte_eth_dev_rss_hash_conf_get. When that
happens, we display whatever that was in the stack variable for rss_hf which
is confusing.
The fix is to clear rss_hf prior to the call.
Before the fix
--------------
DBGvpp# sh hard
Name Idx Link Hardware
format_dpdk_device:599: rte_eth_dev_rss_hash_conf_get returned -95
GigabitEthernet13/0/0 2 up GigabitEthernet13/0/0
Link speed: 10 Gbps
Ethernet address 00:0c:29:c4:8c:eb
VMware VMXNET3
...
rss avail: ipv4 ipv4-tcp ipv6 ipv6-tcp
rss active: ipv4 ipv4-frag ipv4-tcp ipv4-udp ipv4-other
After the fix
-------------
DBGvpp# sh hard
Name Idx Link Hardware
format_dpdk_device:600: rte_eth_dev_rss_hash_conf_get returned -95
GigabitEthernet13/0/0 2 up GigabitEthernet13/0/0
Link speed: 10 Gbps
Ethernet address 00:0c:29:c4:8c:eb
VMware VMXNET3
..
rss avail: ipv4 ipv4-tcp ipv6 ipv6-tcp
rss active: none
Change-Id: If8f8327f0012eecc8d23cd7f3f9cc581ca025654
Signed-off-by: Steven <sluong@cisco.com>
* u32/u64/uword mismatches
* pointer-to-int fixes
* printf formatting issues
* issues with incorrect "ULL" and related suffixes
* structure alignment and padding issues
Change-Id: I70b989007758755fe8211c074f651150680f60b4
Signed-off-by: David Johnson <davijoh3@cisco.com>
fix a copy-paste bug, and a typo of function name
Change-Id: Ib408522d2bb6fde7a7492de6f5d5369b461d77c9
Signed-off-by: Kingwel Xie <kingwel.xie@ericsson.com>
1. multi-loop, and new style with vlib_buffer_enqueue_to_next
2. add error counter for AUTH-FAILURE
3. buffer trace changed. now it supports 'trace add dpdk-crypto-input 10'
just like the other input nodes
Actual measurement shows >10 clocks per packets are saved, under QAT
or openssl PMD case
Change-Id: I6ea34e4ae3b08c381219ff6bc8adda2d927fbfd5
Signed-off-by: Kingwel Xie <kingwel.xie@ericsson.com>
This patch adds support for VMBus to the VPP infrastructure.
Since the only device that matters is the netvsc Poll Mode Driver
in DPDK, the infrastructure is much simpler than PCI.
Change-Id: Ie96c897ad9c426716c2398e4528688ce2217419b
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
otherwise, these pools will occupy an entire huge page for each even
they are very small.
Change-Id: I08919714de9b6cd4b8dddb546ca54364b56ec99f
Signed-off-by: Kingwel Xie <kingwel.xie@ericsson.com>
