Re: [PATCH v2 02/22] PCI: Add API to track PCI devices preserved across Live Update

[David Matlack]

On Tue, Feb 3, 2026 at 4:10 PM Yanjun.Zhu <yanjun.zhu@xxxxxxxxx> wrote:
>
> On 2/2/26 10:14 AM, David Matlack wrote:
> > On Sat, Jan 31, 2026 at 10:38 PM Zhu Yanjun <yanjun.zhu@xxxxxxxxx> wrote:
> >> 在 2026/1/29 13:24, David Matlack 写道:
> >>> Add an API to enable the PCI subsystem to track all devices that are
> >>> preserved across a Live Update, including both incoming devices (passed
> >>> from the previous kernel) and outgoing devices (passed to the next
> >>> kernel).
> >>>
> >>> Use PCI segment number and BDF to keep track of devices across Live
> >>> Update. This means the kernel must keep both identifiers constant across
> >>> a Live Update for any preserved device. VFs are not supported for now,
> >>> since that requires preserving SR-IOV state on the device to ensure the
> >>> same number of VFs appear after kexec and with the same BDFs.
> >>>
> >>> Drivers that preserve devices across Live Update can now register their
> >>> struct liveupdate_file_handler with the PCI subsystem so that the PCI
> >>> subsystem can allocate and manage File-Lifecycle-Bound (FLB) global data
> >>> to track the list of incoming and outgoing preserved devices.
> >>>
> >>>     pci_liveupdate_register_fh(driver_fh)
> >>>     pci_liveupdate_unregister_fh(driver_fh)
> >> Can the above 2 functions support the virtual devices? For example,
> >> bonding, veth, iSWAP and RXE.
> >>
> >> These virtual devices do not have BDF. As such, I am not sure if your
> >> patches take these virtual devices in to account.
> > No this patch series only supports PCI devices, since those are the
> > only devices so far we've needed to support.
> >
> > I am not familiar with any of the devices that you mentioned. If they
> > are virtual then does that mean it's all just software? In that case I
> > would be curious to know what problem is solved by preserving them in
> > the kernel, vs. tearing them down and rebuilding them across a Live
> > Udpate.
>
> Bonding, veth, rxe, and siw can be used in KVM environments.
>
> Although these are software-only virtual devices with no associated
> hardware,
>
> they may maintain state that is observable by userspace.
>
> As a result, Live Update should preserve their state across the update.

Sorry for taking so long to get back to you.

Userspace should serialize the state of those devices out of the
kernel, persist it in memory or on disk across Live Update, and then
recreate it after Live Update.

This is why, for example, KVM does not need to direclty participate in
Live Update. The VMM serializes all KVM state from the kernel, saves
it, and then constructs a completely new KVM VM after the Live Update
using that state. KVM will only need to participate in Live Update
once there is hardware requirement. For example, I believe some of the
Confidential Computing hardware virtualization extensions require
certain memory structures used by hardware remain in place and cannot
be reallocated after Live Update. KVM would have to participate to
preserve those and reassociate them with the right VM after Live
Update.