| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
We already had /if/lacp/<bond-sw_if_index>/<slave-sw_if_index>/state in
the stats segment. Add also the partner-state to be complete.
Change to populate stats segment with the states at startup, after processing
an lacp pdu, and after timer expiration.
Unit test
---------
sudo vpp_get_stats dump | grep /if/lacp
63.00 /if/lacp/3/1/state
61.00 /if/lacp/3/1/partner-state
63.00 /if/lacp/3/2/state
61.00 /if/lacp/3/2/partner-state
Type: feature
Signed-off-by: Steven Luong <sluong@cisco.com>
Change-Id: Ib7b8e1183d572bb6e422a846aaa2b7b3559a0dc7
|
|
In a rare event, we may be skipping processing lacp pdu's when the it is
not in steady state.
Type: fix
Signed-off-by: Steven Luong <sluong@cisco.com>
Change-Id: I3595d22dbff8a97dce9fb4d4452d2051bcf6f523
|
|
Not all interfaces have the same characteristics within the bonding group.
For active-backup mode, we should do our best to select the slave that
performs the best as the primary slave. We already did that by preferring
the slave that is local numa. Sometimes, this is not enough. For example,
when all are local numas, the selection is arbitrary. Some slave interfaces
may have higher speed or better qos than the others. But this is hard to
infer.
One rule does not fit all. So we let the operator to optionally specify the
weight for each slave interface. Our primary slave selection rule is now
1. biggest weight
2. is local numa
3. current primary slave (to avoid churn)
4. lowest sw_if_index (for deterministic behavior)
This selection rule only applies to active-backup mode which only one slave
is used for forwarding traffic until it becomes unreachable. At that time,
the next "best" slave candidate is automatically promoted. The slaves are
sorted according to the preference rule when they are up. So there is no need
to find the next best candidate when the primary slave goes down.
Another good thing about this rule is when the down slave comes back up, it
is selected as the primary slave again unless there is indeed a "better"
slave than this down slave that were added during that period.
To set the weight for the slave interface, do this after the interface is
enslaved
set interface bond <interface-name> weight <value>
Type: feature
Signed-off-by: Steven Luong <sluong@cisco.com>
Change-Id: I59ced6d20ce1dec532e667dbe1afd1b4243e04f9
|
|
Add /if/lacp/<bond-sw_if_index>/<slave-sw_if_index>/state
<bond-sw_if_index> is a vector of the bond sw_if_index
<slave-sw_if_index> is a vector of the slave sw_if_index
Content is the integer value of the lacp actor state. The state is actually
a bitfield as described in the lacp protocol spec.
Type: feature
Signed-off-by: Steven Luong <sluong@cisco.com>
Change-Id: Ic6eca8ce2a1acd2d858e4e50b7eac1d000ea08e5
Signed-off-by: Ole Troan <ot@cisco.com>
|
|
If numa-only is set, Only slaves on local numa node
transmit pkts if have at least one, otherwise the bond
interface works as usual.
CLI change:
create bond mode lacp [load-balance { l2 | l23 | l34 } {numa-only}]
[hw-addr <mac-address>] [id <if-id>]
The new member "u8 numa_only;" is also added to bond_create_if_args_t.
Type: feature
Change-Id: Icdccedafb0738d8c9d4a5acce909ce562428c071
Signed-off-by: Zhiyong Yang <zhiyong.yang@intel.com>
|
|
Type: style
Change-Id: I28908756019f8ca54c50334c470d8eded5621ade
Signed-off-by: Zhiyong Yang <zhiyong.yang@intel.com>
|
|
This patch enables bonding numa awareness on multi-socket
server working in active-backeup mode.
The VPP adds capability for automatically preferring slave
with local numa node in order to reduces the load on the
QPI-bus and improve system overall performance in multi-socket
use cases. Users doesn't need to add any extra operation as
usual.
Change-Id: Iec267375fc399a9a0c0a7dca649fadb994d36671
Signed-off-by: Zhiyong Yang <zhiyong.yang@intel.com>
|
|
By definition, passive mode means the node does not start sending lacp pdu until
it first hears from the partner or remote.
- Rename ptx machine's BEGIN state to NO_PERIODIC state.
- Put periodic machine in NO_PERIDOIC state when the interface is enabled for
lacp. ptx machine will transition out of NO_PERIODIC state when the local node
hears from the remote or when the local node is configured for active mode.
- Also add send and receive statistics for debugging.
Change-Id: I747953b9595ed31328b2f4f3e7a8d15d01e04d7f
Signed-off-by: Steven Luong <sluong@cisco.com>
|
|
-fno-common makes sure we do not have multiple declarations of the same
global symbol across compilation units. It helps debug nasty linkage
bugs by guaranteeing that all reference to a global symbol use the same
underlying object.
It also helps avoiding benign mistakes such as declaring enum as global
objects instead of types in headers (hence the minor fixes scattered
across the source).
Change-Id: I55c16406dc54ff8a6860238b90ca990fa6b179f1
Signed-off-by: Benoît Ganne <bganne@cisco.com>
|
|
Change-Id: I78fe58144fa3ba2e1c7135897a13a2541f235c91
Signed-off-by: Alexander Chernavin <achernavin@netgate.com>
|
|
Break up bond tx function into multiple small workloads:
1. parse the packet header and hash it based on the configured algorithm
2. optionally, trace the packet
3. convert the hash value from (1) to the slave port
4. update the buffers with the slave sw_if_index
5. Add the buffers to the queues
6. Create and send the frames
old numbers
-----------
Time 5.3, average vectors/node 223.74, last 128 main loops 40.00 per node 222.61
vector rates in 3.3627e6, out 6.6574e6, drop 3.3964e4, punt 0.0000e0
Name State Calls Vectors Suspends Clocks Vectors/Call
BondEthernet0-output active 68998 17662979 0 1.89e1 255.99
BondEthernet0-tx active 68998 17662979 0 2.60e1 255.99
TenGigabitEthernet3/0/1-output active 68998 8797416 0 1.03e1 127.50
TenGigabitEthernet3/0/1-tx active 68998 8797416 0 7.85e1 127.50
TenGigabitEthernet7/0/1-output active 68996 8865563 0 1.02e1 128.49
TenGigabitEthernet7/0/1-tx active 68996 8865563 0 7.65e1 128.49
new numbers
-----------
BondEthernet0-output active 304064 77840384 0 2.29e1 256.00
BondEthernet0-tx active 304064 77840384 0 2.47e1 256.00
TenGigabitEthernet3/0/1-output active 304064 38765525 0 1.03e1 127.49
TenGigabitEthernet3/0/1-tx active 304064 38765525 0 7.66e1 127.49
TenGigabitEthernet7/0/1-output active 304064 39074859 0 1.01e1 128.51
Change-Id: I3ef9a52bfe235559dae09d055c03c5612c08a0f7
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
Introduce bond_tx_inline which takes lb as a constant for gcc to do the optimization
The number appears a tad better for 256 bytes frame.
with the patch
--------------
Thread 2 vpp_wk_1 (lcore 3)
Time 4.3, average vectors/node 224.00, last 128 main loops 40.00 per node 222.61
vector rates in 8.4836e6, out 1.6967e7, drop 0.0000e0, punt 0.0000e0
Name State Calls Vectors Suspends Clocks Vectors/Call
BondEthernet0-output active 141054 36109824 0 2.51e1 256.00
BondEthernet0-tx active 141054 36109824 0 2.55e1 256.00
TenGigabitEthernet6/0/0-output active 141054 18055469 0 9.43e0 128.00
TenGigabitEthernet6/0/0-tx active 141054 18055469 0 6.97e1 128.00
TenGigabitEthernet6/0/1-output active 141054 18054355 0 9.54e0 127.99
TenGigabitEthernet6/0/1-tx active 141054 18054355 0 7.05e1 127.99
bond-input active 141054 36109824 0 1.76e1 256.00
dpdk-input polling 70527 36109824 0 5.03e1 512.00
ethernet-input active 141054 36109824 0 6.12e1 256.00
ip4-input active 141054 36109824 0 3.26e1 256.00
ip4-lookup active 141054 36109824 0 2.94e1 256.00
ip4-rewrite active 141054 36109824 0 3.27e1 256.00
without the patch
-----------------
Thread 2 vpp_wk_1 (lcore 3)
Time 4.3, average vectors/node 224.00, last 128 main loops 40.00 per node 222.61
vector rates in 8.4443e6, out 1.6889e7, drop 0.0000e0, punt 0.0000e0
Name State Calls Vectors Suspends Clocks Vectors/Call
BondEthernet0-output active 142744 36542464 0 2.51e1 256.00
BondEthernet0-tx active 142744 36542464 0 2.67e1 256.00
TenGigabitEthernet6/0/0-output active 142744 18270813 0 9.19e0 127.99
TenGigabitEthernet6/0/0-tx active 142744 18270813 0 6.98e1 127.99
TenGigabitEthernet6/0/1-output active 142744 18271651 0 9.43e0 128.00
TenGigabitEthernet6/0/1-tx active 142744 18271651 0 7.02e1 128.00
bond-input active 142744 36542464 0 1.76e1 256.00
dpdk-input polling 71372 36542464 0 5.08e1 512.00
ethernet-input active 142744 36542464 0 6.15e1 256.00
ip4-input active 142744 36542464 0 3.23e1 256.00
ip4-lookup active 142744 36542464 0 2.96e1 256.00
ip4-rewrite active 142744 36542464 0 3.28e1 256.00
Change-Id: I9fd43eda3c735cbff680ac6d2f01ecdae81f0eda
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
- Reduce per packet cost by buffering the output packet buffer indexes in the queue and
process the queue outside the packet processing loop.
- Move unnecessary variable initialization outside of the while loop.
- There is no need to save the old interface if tracing is not enabled.
Test result for 256 bytes packet comparison. Other packet size shows similar improvement.
With the patch
--------------
BondEthernet0-output active 52836 13526016 0 1.71e1 256.00
BondEthernet0-tx active 52836 13526016 0 2.68e1 256.00
TenGigabitEthernet6/0/0-output active 52836 6762896 0 9.17e0 127.99
TenGigabitEthernet6/0/0-tx active 52836 6762896 0 6.97e1 127.99
TenGigabitEthernet6/0/1-output active 52836 6763120 0 9.40e0 128.00
TenGigabitEthernet6/0/1-tx active 52836 6763120 0 7.00e1 128.00
bond-input active 52836 13526016 0 1.76e1 256.00
Without the patch
-----------------
BondEthernet0-output active 60858 15579648 0 1.73e1 256.00
BondEthernet0-tx active 60858 15579648 0 2.94e1 256.00
TenGigabitEthernet6/0/0-output active 60858 7789626 0 9.29e0 127.99
TenGigabitEthernet6/0/0-tx active 60858 7789626 0 7.01e1 127.99
TenGigabitEthernet6/0/1-output active 60858 7790022 0 9.31e0 128.00
TenGigabitEthernet6/0/1-tx active 60858 7790022 0 7.10e1 128.00
bond-input active 60858 15579648 0 1.77e1 256.00
Change-Id: Ib6d73a63ceeaa2f1397ceaf4c5391c57fd865b04
Signed-off-by: Steven <sluong@cisco.com>
|
|
Change-Id: I0c3f2add35ad9fc11308b7a2a2c69ffd8472dd2e
Signed-off-by: Damjan Marion <damarion@cisco.com>
|
|
- Modify the API send_ip6_na and send_ip4_garp to take sw_if_index instead
of vnet_hw_interface_t and add call to build_ethernet_rewrite to support
subinterface/vlan
- Add code to bonding driver to send an event to bond_process when the first
interface becomes active or when the active interface is down
- Create a bond_process to walk the interface and the corresponding
subinterfaces to send garp/ip6_na when an event is received.
- Minor cleanup in bonding/node.c
Note: dpdk bonding driver does not send garp/ip6_na for subinterfaces. There is
no attempt to fix it here. But the infra is now done and should be easy to
add the support.
Change-Id: If3ecc4cd0fb3051330f7fa11ca0dab3e18557ce1
Signed-off-by: Steven <sluong@cisco.com>
|
|
- hash is great. But it is a bit too slow for the DP. Use direct array indexing
to quickly retrieve the slave interface.
- the algorithm used by flow hash is great. But it is a bit too slow for the DP.
Use l2_hash_hash() extracted from lb_hash.h which ECMP is using. It makes use
of intrinsic crc32 instruction set.
- shortcut modulo arithmetic when the operand is 2**x (where x up to 4) to
avoid division instruction.
- special case for link count == 1 in bond_tx_fn()
- use clib_mem_unaligned to access data for the packet to avoid alignment error
- Fix some typos for packet tracing.
Change-Id: I8eae3ad497061c5473aa675ba894ee0211120d25
Signed-off-by: Steven <sluong@cisco.com>
|
|
coverity complains about statements in function A
function A
{
x % vec_len (y)
}
because vec_len (y) is a macro and may return 0 if the pointer y is null.
But coverity fails to realize the same statement vec_len (y) was already
invoked and checked in the caller of function A and punt if vec_len (y) is 0.
We can fix the coverity warning and shave off a few cpu cycles by caching
the result of vec_len (y) and pass it around to avoid calling vec_len (y)
again in multiple places.
Change-Id: I095166373abd3af3859646f860ee97c52f12fb50
Signed-off-by: Steven <sluong@cisco.com>
|
|
We were only puting one packet per frame to the output node. Change to
buffer multiple packets per frame. Performance is now on top of dpdk-based
bonding.
Put a spinlock in the tx thread in case the rug is pulled under us.
Change-Id: Ifda5af086a984a7301972cd6c8e428217f676a95
Signed-off-by: Steven <sluong@cisco.com>
|
|
Add bonding driver to support creation of bond interface which composes of
multiple slave interfaces. The slave interfaces could be physical interfaces,
or just any virtual interfaces. For example, memif interfaces.
The syntax to create a bond interface is
create bond mode <lacp | xor | acitve-backup | broadcast | round-robin>
To enslave an interface to the bond interface,
enslave interface TenGigabitEthernet6/0/0 to BondEthernet0
Please see src/plugins/lacp/lacp_doc.md for more examples and additional
options.
LACP is a control plane protocol which manages and monitors the status of
the slave interfaces. The protocol is part of 802.3ad standard. This patch
implements LACPv1. LACPv2 is not supported.
To enable LACP on the bond interface, specify "mode lacp" when the bond
interface is created. The syntax to enslave a slave interface is the same as
other bonding modes.
Change-Id: I06581d3b87635972f9f0e1ec50b67560fc13e26c
Signed-off-by: Steven <sluong@cisco.com>
|
Steven Luong
Zhiyong Yang
Benoît Ganne
Alexander Chernavin
Damjan Marion