4.15.0-1037 does not see all PCI devices on GPU VMs
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
linux-azure (Ubuntu) |
Fix Released
|
Undecided
|
Unassigned | ||
Xenial |
Fix Released
|
Undecided
|
Marcelo Cerri | ||
Cosmic |
Fix Released
|
Undecided
|
Marcelo Cerri |
Bug Description
Host changes have altered how the PCI GUID is presented to the guest and the patches for PCI IDs in 4.15.0-1037 do not properly handle the new condition.
Impact:
Instances with multiple GPUs are only seeing one.
Workaround:
4.15.0-1036 does not have this behavior.
Additional info:
The commit in 4.15.0-1037 responsible is " - PCI: hv: Make sure the bus domain is really unique"
The immediate action requested is to back out this patch on 4.15.0 (azure):
https:/
The same thing on 4.18:
https:/
no longer affects: | linux-azure (Ubuntu Bionic) |
Changed in linux-azure (Ubuntu Xenial): | |
status: | New → Fix Released |
Changed in linux-azure (Ubuntu Cosmic): | |
status: | New → Fix Released |
Changed in linux-azure (Ubuntu Xenial): | |
assignee: | nobody → Marcelo Cerri (mhcerri) |
Changed in linux-azure (Ubuntu Cosmic): | |
assignee: | nobody → Marcelo Cerri (mhcerri) |
This bug was fixed in the package linux-azure - 4.18.0-1011.11
---------------
linux-azure (4.18.0-1011.11) cosmic; urgency=medium
* linux-azure: 4.18.0-1011.11 -proposed tracker (LP: #1816081)
* 4.15.0-1037 does not see all PCI devices on GPU VMs (LP: #1816106)
- Revert "PCI: hv: Make sure the bus domain is really unique"
linux-azure (4.18.0-1009.9) cosmic; urgency=medium
* Allow I/O schedulers to be loaded with modprobe in linux-azure
(LP: #1813211)
- [Config] linux-azure: Enable all IO schedulers as modules
* [Hyper-V] srcu: Lock srcu_data structure in srcu_gp_start() (LP: #1802021)
- srcu: Lock srcu_data structure in srcu_gp_start()
* CONFIG_ SECURITY_ SELINUX_ DISABLE should be disabled on 4.15/4.18 Azure SECURITY_ SELINUX_ DISABLE
(LP: #1813866)
- [Config]: disable CONFIG_
[ Ubuntu: 4.18.0-15.16 ]
* Ubuntu boot failure. 4.18.0-14 boot stalls. (does not boot) (LP: #1814555) ringbuffer: Delay after EMIT_INVALIDATE for gen4/gen5"
- Revert "drm/i915/
* Userspace break as a result of missing patch backport (LP: #1813873)
- tty: Don't hold ldisc lock in tty_reopen() if ldisc present
-- Stefan Bader <email address hidden> Fri, 15 Feb 2019 17:16:24 +0100