RE: [PATCH 1/1] Documentation: hyperv: Add overview of guest VM hibernation
From: Michael Kelley <hidden>
Date: 2024-12-13 20:43:42
Also in:
linux-doc, lkml
From: Roman Kisel <redacted> Sent: Friday, December 13, 2024 10:44 AM
On 12/12/2024 3:17 PM, mhkelley58@gmail.com wrote:quoted
From: Michael Kelley <redacted> Add documentation on how hibernation works in a guest VM on Hyper-V. Describe how VMBus devices and the VMBus itself are hibernated and resumed, along with various limitations.
[snip]
quoted
+Considerations for Guest VM Hibernation +--------------------------------------- +Linux guests on Hyper-V can also be hibernated, in which case the +hardware is the virtual hardware provided by Hyper-V to the guest VM. +Only the targeted guest VM is hibernated, while other guest VMs and +the underlying Hyper-V host continue to run normally. While the +underlying Windows Hyper-V and physical hardware on which it is +running might also be hibernated using hibernation functionality in +the Windows host, host hibernation and its impact on guest VMs is not +in scope for this documentation. + +Resuming a hibernated guest VM can be more challenging than with +physical hardware because VMs make it very easy to change the hardware +configuration between the hibernation and resume. Even when the resume +is done on the same VM that hibernated, the memory size might be +changed, or virtual NICs or SCSI controllers might be added or +removed. Virtual PCI devices assigned to the VM might be added or +removed. Most such changes cause the resume steps to fail, though +adding a new virtual NIC, SCSI controller, or vPCI device should work. +Would it be useful mentioning the (likely lethal for the VM) risk of copying the hibernated VM to another host (of the same arch) that has another set of CPUID bits/features?
Yes, that's a good point that is specific to VMs. I'll add it to the documentation. [snip]
Appreciated documenting all the intricacies of the hibernation and resume paths for various devices, an incredible read! Are there any special considerations known to you for the hibernation of the devices driven through the Hyper-V UIO?
The UIO driver for VMBus devices (uio_hv_generic.c) does not have support for hibernation -- it does not have "suspend" and "resume" functions implemented like the other VMBus device drivers. Consequently, vmbus_suspend() returns an EOPNOTSUPP error (-95) when Linux goes through the hibernation sequence. The error causes the sequence to abort, and the VM is not hibernated. FWIW, here's example output: [86945.335293] PM: hibernation: hibernation entry [86945.344403] Filesystems sync: 0.008 seconds [86945.344853] Freezing user space processes [86945.346331] Freezing user space processes completed (elapsed 0.001 seconds) [86945.346340] OOM killer disabled. [86945.346410] PM: hibernation: Marking nosave pages: [mem 0x00000000-0x00000fff] [86945.346412] PM: hibernation: Marking nosave pages: [mem 0x000a0000-0x000fffff] [86945.346413] PM: hibernation: Marking nosave pages: [mem 0xee9b4000-0xee9bafff] [86945.346414] PM: hibernation: Marking nosave pages: [mem 0xeff41000-0xeffc4fff] [86945.346415] PM: hibernation: Marking nosave pages: [mem 0xeffd3000-0xefffefff] [86945.346416] PM: hibernation: Marking nosave pages: [mem 0xf0000000-0xffffffff] [86945.346649] PM: hibernation: Basic memory bitmaps created [86945.346659] PM: hibernation: Preallocating image memory [86946.173088] PM: hibernation: Allocated 611162 pages for snapshot [86946.173105] PM: hibernation: Allocated 2444648 kbytes in 0.82 seconds (2981.27 MB/s) [86946.173114] Freezing remaining freezable tasks [86946.174472] Freezing remaining freezable tasks completed (elapsed 0.001 seconds) [86946.174568] printk: Suspending console(s) (use no_console_suspend to debug) [86946.205965] uio_hv_generic 1fe3f3ee-76ed-4bfe-871f-984d1563c7a2: PM: dpm_run_callback(): vmbus_suspend [hv_vmbus] returns -95 [86946.205989] uio_hv_generic 1fe3f3ee-76ed-4bfe-871f-984d1563c7a2: PM: failed to freeze noirq: error -95 [86946.206812] PM: hibernation: Some devices failed to power down, aborting [86946.246700] PM: hibernation: Basic memory bitmaps freed [86946.247759] OOM killer enabled. [86946.247766] Restarting tasks ... done. [86946.249415] PM: hibernation: hibernation exit I'll add this limitation to the documentation as well. Given the plans to use the UIO driver for a broader set of specialty VMBus devices, this is a limitation that likely needs to be remedied. Thanks for the input! These are useful points that I had not considered. Michael