xCAT/xcat-core
2
0
Fork 0
mirror of https://github.com/xcat2/xcat-core.git synced 2025-07-27 23:01:12 +00:00
Code Issues Releases Wiki Activity

Add new PCI format support for pcipassthrough (#3916)

Browse Source 
* add devpassthrough support

* add new pci format for pcipassthrough

* add new pci format support

* polished

* update doc
This commit is contained in:
Yuan Bai
2017-09-14 17:15:47 +08:00
committed by yangsong
parent 2e403a3e8c
commit d45f7d4508
3 changed files with 31 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
docs/source/guides/admin-guides/manage_clusters/common/kvm/manage_vm.rst
View File

@@ -82,7 +82,27 @@ After the VM object is created, several key attributes need to be specified with
chtab node=vm1 vm.vidpassword=abc123
10. Set **netboot** attribute
10. (optional)For assigning PCI devices to the VM, set **othersettings** value: ::
chtab node=vm1 vm.othersettings="devpassthrough:0000:01:00.2"
Or: ::
chtab node=vm1 vm.othersettings="devpassthrough:pci_0000_01_00_2"
Take assigning SR-IOV VFs to the VM as an example:
* Use ``lspci`` to get VFs PCI from hypervisor: ::
lspci|grep -i "Virtual Function"
0000:01:00.1 Infiniband controller: Mellanox Technologies MT27700 Family [ConnectX-4 Virtual Function]
0000:01:00.2 Infiniband controller: Mellanox Technologies MT27700 Family [ConnectX-4 Virtual Function]
* Set the VFs PCI into ``vm`` table on MN: ::
chtab node=vm1 vm.othersettings="devpassthrough:0000:01:00.1,0000:01:00.2"
11. Set **netboot** attribute
* **[x86_64]** ::

2
perl-xCAT/xCAT/Schema.pm
View File

@@ -226,7 +226,7 @@ qq{ link,ro - The file is readonly, and will be placed in tmpfs on the booted no
'vncport' => 'Tracks the current VNC display port (currently not meant to be set',
'textconsole' => 'Tracks the Psuedo-TTY that maps to the serial port or console of a VM',
'powerstate' => "This flag is used by xCAT to track the last known power state of the VM.",
'othersettings' => "This allows specifying a semicolon delimited list of key->value pairs to include in a vmx file of VMware or KVM. For partitioning on normal power machines, this option is used to specify the hugepage and/or bsr information, the value is like:'hugepage:1,bsr=2'. For KVM cpu pinning, this option is used to specify the physical cpu set on the host, the value is like:\"vcpupin:'0-15,^8'\",Its syntax is a comma separated list and a special markup using '-' and '^' (ex. '0-4', '0-3,^2') can also be allowed, the '-' denotes the range and the '^' denotes exclusive. For KVM memory binding, the value is like:'membind:0', restrict a guest to allocate memory from the specified set of NUMA nodes. For PCI passthrough, the value is like:'devpassthrough:pci_0001_01_00_0,pci_0000_03_00_0',the PCI devices are assigned to a virtual machine, and the virtual machine can use this I/O exclusively, the devices list are a list of PCI device names delimited with comma, the PCI device names can be obtained by running B<virsh nodedev-list> on the host.",
'othersettings' => "This allows specifying a semicolon delimited list of key->value pairs to include in a vmx file of VMware or KVM. For partitioning on normal power machines, this option is used to specify the hugepage and/or bsr information, the value is like:'hugepage:1,bsr=2'. For KVM cpu pinning, this option is used to specify the physical cpu set on the host, the value is like:\"vcpupin:'0-15,^8'\",Its syntax is a comma separated list and a special markup using '-' and '^' (ex. '0-4', '0-3,^2') can also be allowed, the '-' denotes the range and the '^' denotes exclusive. For KVM memory binding, the value is like:'membind:0', restrict a guest to allocate memory from the specified set of NUMA nodes. For PCI passthrough, the value is like:'devpassthrough:pci_0001_01_00_0,pci_0000_03_00_0', the value for PCI device format also can be like:'devpassthrough:0001:01:00.1', the PCI devices are assigned to a virtual machine, and the virtual machine can use this I/O exclusively, the devices list are a list of PCI device names delimited with comma, the PCI device names can be obtained by running B<virsh nodedev-list> on the host.",
'guestostype' => "This allows administrator to specify an identifier for OS to pass through to virtualization stack. Normally this should be ignored as xCAT will translate from nodetype.os rather than requiring this field be used\n",
'beacon' => "This flag is used by xCAT to track the state of the identify LED with respect to the VM.",
'datacenter' => "Optionally specify a datacenter for the VM to exist in (only applicable to VMWare)",

9
xCAT-server/lib/xcat/plugins/kvm.pm
View File

@@ -783,6 +783,15 @@ sub build_xmldesc {
#prepare the xml hash for pci passthrough
my @prdevarray;
foreach my $devname (@passthrudevices) {
#This is for SR-IOV vfio
#Change vfio format 0000:01:00.2 to pci_0000_01_00_2
if ( $devname =~ m/(\w:)+(\w)+.(\w)/ ){
$devname =~ s/[:|.]/_/g;
if ( $devname !~ /^pci_/ ) {
$devname ="pci_".$devname
}
}
my $devobj = $hypconn->get_node_device_by_name($devname);
unless ($devobj) {
return -1;