vmm: add explicit PCI BDF support for Rng device - #8163
Conversation
| @@ -1750,23 +1750,25 @@ impl NetConfig { | |||
| } | |||
|
|
|||
| impl RngConfig { | |||
| pub const SYNTAX: &'static str = | |||
| "Random number generator parameters \"src=<path:/dev/urandom>,pci_device_id=<pci_slot>\""; | |||
There was a problem hiding this comment.
open question. should I advertise pci_segment here or not? Technically it is supported by the parsing logic but I am not sure if it would work.
We (@cyberus-technology) do not need support for this. We just need pci_device_id
2a0c74a to
c97e82e
Compare
|
@rbradford it would be lovely if we can get this merged before the upcoming v52 release. I think it is fairly uncontroversial and a nice fix/feature |
rbradford
left a comment
There was a problem hiding this comment.
lgtm, We can definitely land this soon.
| } | ||
|
|
||
| impl RngConfig { | ||
| // We ignore `id` and `pci_segment` as we use hard-coded internal defaults for now. |
There was a problem hiding this comment.
Is this comment still accurate?
There was a problem hiding this comment.
Yikes, will remove it. Thank!
|
Sorry @phip1611 i'd missed you'd updated this PR today. I wonder if GH is having notification issues. |
This was missing in [0] but is required for proper explicit PCI BDF management, e.g., when a VM is created via libvirt and each device has an explicit BDF. [0] cloud-hypervisor#7965 On-behalf-of: SAP [email protected] Signed-off-by: Philipp Schuster <[email protected]>

This was missing in [0] but is required for proper explicit PCI BDF management, e.g., when a VM is created via libvirt and each device has an explicit BDF. Also important for live migration.
Without this, the BDF could change after a live migration, due to implicit BDF allocation! Further, management software wants full control over PCI BDFs.
[0] #7965