aboutsummaryrefslogtreecommitdiffstats
path: root/src/plugins/avf/README.md
blob: 40e4f0cd04028730a9d2f067e10041297d45489b (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
# Intel AVF device plugin for VPP    {#avf_plugin_doc}

##Overview
This plugins provides native device support for intel Adaptive Virtual
Function (AVF). AVF is driver specification for current and future
Intel Virtual Function devices. AVF defines communication channel between
Physical Functions (PF) and VF.
In essence, today this driver can be used only with 
Intel XL710 / X710 / XXV710 adapters.

##Prerequisites
 * Driver requires newer i40e PF linux driver to be installed on the system,
which supports virtualchnl interface. This code is tested with i40e driver
version 2.4.6.

* Driver requires MSI-X interrupt support, which is not supported by
uio_pci_generic driver, so vfio-pci needs to be used. On systems without IOMMU
vfio driver can still be used with recent kernels which support no-iommu mode.

##Known issues
This driver is still in experimental phase, however it shows very good 
performance numbers. Following items are not implemented (yet).

* Jumbo MTU support
* Adaptive mode
* NUMA support

## Usage
### System setup

1. load VFIO driver
```
sudo modprobe vfio-pci
```

2. (systems without IOMMU only) enable unsafe NOIOMMU mode
```
echo Y | sudo tee /sys/module/vfio/parameters/enable_unsafe_noiommu_mode
```

3. Create and bind SR-IOV virtual function(s)

Following script creates VF, assigns MAC address and binds VF to vfio-pci
```bash
#!/bin/bash

if [ $USER != "root" ] ; then
	echo "Restarting script with sudo..."
	sudo $0 ${*}
	exit
fi

setup () {
  cd /sys/bus/pci/devices/${1}
  driver=$(basename $(readlink driver))
  if [ "${driver}" != "i40e" ]; then
    echo ${1} | tee driver/unbind
    echo ${1} | tee /sys/bus/pci/drivers/i40e/bind
  fi
  ifname=$(basename net/*)
  echo 0 | tee sriov_numvfs > /dev/null
  echo 1 | tee sriov_numvfs > /dev/null
  ip link set dev ${ifname} vf 0 mac ${2}
  ip link show dev ${ifname}
  vf=$(basename $(readlink virtfn0))
  echo ${vf} | tee virtfn0/driver/unbind
  echo vfio-pci | tee virtfn0/driver_override
  echo ${vf} | sudo tee /sys/bus/pci/drivers/vfio-pci/bind
  echo  | tee virtfn0/driver_override
}

# Setup one VF on PF 0000:3b:00.0 and assign MAC address
setup 0000:3b:00.0 00:11:22:33:44:00
# Setup one VF on PF 0000:3b:00.1 and assign MAC address
setup 0000:3b:00.1 00:11:22:33:44:01
```

### Promisc mode
In cases when interface is used in the L2 mode or promisc mode is needed for some other reason,
trust needs to be set to "on" using the linux "ip link" utility.
```
ip link set dev <PF inteface name> vf <VF id> trust on
```

### L2 spoofing check
By default Virtual Function is not allowed to send ethernet frames which
have source MAC address different than address assigned to the VF.
In some cases it is expected that VPP will send such frames (e.g. L2 bridging,
bonding, l2 cross-connect) and in such cases spoof chack needs to be turned
off by issuing following command:
```
ip link set dev <PF inteface name> vf <VF id> spoofchk off
```

### Interface Creation
Interfaces can be dynamically created by using following CLI:
```
create interface avf 0000:3b:02.0
set int state avf-0/3b/2/0 up
```

### Interface Deletion
Interface can be deleted with following CLI:
```
delete interface avf <interface name>
```

### Interface Statistics
Interface statistics can be displayed with `sh hardware-interface <if-name>`
command.