It was a bit strange that the jammy 5.15 kernel worked on the v6 instance type,
so I also went and verified for 6.8 on jammy HWE.

Performing verification for jammy HWE.

I created two VMs on Azure, one of each:
- Standard_D4ads_v5
- Standard_D4ads_v6

using the - "Ubuntu Server 22.04 LTS - x64 Gen2" image.

On each instance I turned off Secure Boot.

Each VM is running 6.8.0-1062-azure from -updates.

$ uname -rv
6.8.0-1062-azure #69~22.04.1-Ubuntu SMP Mon Jun 29 19:32:04 UTC 2026

I installed kdump-tools, configured the modules to include in the initramfs, and
rebooted.

From there I opened the serial console on Azure, and crashed the VM with sysrq
trigger.

The v6 instance gets stuck during kexec boot. Looking through the serial console
I see:

[    0.769578] hv_vmbus: registering driver hv_pci
[    0.775870] hv_pci 7ad35d50-c05b-47ab-b3a0-56a9a845852b: PCI VMBus probing: 
Using version 0x10004
[    0.785099] hv_pci 7ad35d50-c05b-47ab-b3a0-56a9a845852b: Retrying D0 Entry
[    0.790771] hv_pci 7ad35d50-c05b-47ab-b3a0-56a9a845852b: PCI Pass-through 
VSP failed D0 Entry with status c0370048
[    0.800755] hv_vmbus: probe failed for device 
7ad35d50-c05b-47ab-b3a0-56a9a845852b (-71)
[    0.804779] hv_pci: probe of 7ad35d50-c05b-47ab-b3a0-56a9a845852b failed 
with error -71
[    0.819251] hv_pci 4fa6bdc7-d125-4935-9ade-79c121171051: PCI VMBus probing: 
Using version 0x10004
[    0.828004] hv_pci 4fa6bdc7-d125-4935-9ade-79c121171051: Retrying D0 Entry
[    0.850997] hv_pci 4fa6bdc7-d125-4935-9ade-79c121171051: PCI Pass-through 
VSP failed D0 Entry with status c0370048
[    0.862687] hv_vmbus: probe failed for device 
4fa6bdc7-d125-4935-9ade-79c121171051 (-71)
[    0.866584] hv_pci: probe of 4fa6bdc7-d125-4935-9ade-79c121171051 failed 
with error -71
[    0.881378] hv_pci 7b105564-b004-45ec-aba6-6637afa95dde: PCI VMBus probing: 
Using version 0x10004
[    0.891098] hv_pci 7b105564-b004-45ec-aba6-6637afa95dde: Retrying D0 Entry
[    0.897780] hv_pci 7b105564-b004-45ec-aba6-6637afa95dde: PCI Pass-through 
VSP failed D0 Entry with status c0370048
[    0.908555] hv_vmbus: probe failed for device 
7b105564-b004-45ec-aba6-6637afa95dde (-71)
[    0.912513] hv_pci: probe of 7b105564-b004-45ec-aba6-6637afa95dde failed 
with error -71
[    0.919503] Freeing initrd memory: 32556K

and also

[    1.470430] hyperv_drm 5620e0c7-8062-4dce-aeb7-520c7ef76171: [drm] Synthvid 
Version major 3, minor 5
[    1.472227] hv_vmbus: registering driver hv_netvsc
[    1.472906] hv_vmbus: registering driver hyperv_keyboard
[    1.473890] [drm] Initialized hyperv_drm 1.0.0 2020 for 
5620e0c7-8062-4dce-aeb7-520c7ef76171 on minor 0
[    1.479156] Console: switching to colour frame buffer device 128x48
[    1.479656] input: AT Translated Set 2 keyboard as 
/devices/LNXSYSTM:00/LNXSYBUS:00/ACPI0004:00/MSFT1000:00/d34b2567-b9b6-42b9-8778-0a4ec0b955bf/serio0/input/input0
[    1.485556] hyperv_drm 5620e0c7-8062-4dce-aeb7-520c7ef76171: [drm] fb0: 
hyperv_drmdrmfb frame buffer device
[    1.488192] hv_vmbus: registering driver hid_hyperv
[    1.495582] input: Microsoft Vmbus HID-compliant Mouse as 
/devices/0006:045E:0621.0001/input/input1
[    1.497750] hid-hyperv 0006:045E:0621.0001: input: VIRTUAL HID v0.01 Mouse 
[Microsoft Vmbus HID-compliant Mouse] on 
[    1.508783] hv_netvsc 002248e2-f81d-0022-48e2-f81d002248e2 (unnamed 
net_device) (uninitialized): VF slot 1 added
Begin: Loading essential drivers ... [    1.627869] rdac: device handler 
registered

It seems that hyperv_drm wins access to the MMIO range, and hv_pci gets left out
in the cold. Which tracks, because I get no rootfs, and no kdump is captured.

When I log into the system, there is no kdump in /var/crash.

The v5 system correctly captures the kdump.

I then enabled the kernel team ppa, since linux-azure-6.8 is not in -proposed 
yet,
and installed the 6.8.0-1063-azure kernel and rebooted.

$ uname -rv
6.8.0-1063-azure #71~22.04.1-Ubuntu SMP Mon Jul  6 22:48:31 UTC 2026

From there I repeated the same crashdump collection.

Both the v5 and v6 instances correctly collected a crashdump and wrote it to
/var/crash.

There does not appear to be any regressions on v5. It solves the problem on v6,
and v6 can crashdump correctly now. Marking verified for jammy HWE.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2155222

Title:
  [hyperv] Ensure MMIO Mapping is Correct for Kexec / kdump kernel on
  Azure v6 Instance Types

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux-azure/+bug/2155222/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to