vmm: restore KVM clock before resuming vCPUs - #7932
Merged
rbradford merged 1 commit intoMar 31, 2026
Merged
Conversation
phip1611
reviewed
Mar 31, 2026
phip1611
left a comment
Member
There was a problem hiding this comment.
Thanks! Let me run this through our pipeline.
LGTM but I want to make sure it's working reliably
CMGS
force-pushed
the
fix/resume-clock-ordering
branch
from
March 31, 2026 06:27
c2af4bd to
c2324b3
Compare
phip1611
reviewed
Mar 31, 2026
phip1611
approved these changes
Mar 31, 2026
phip1611
left a comment
Member
There was a problem hiding this comment.
LGTM! Please address the remaining remarks
phip1611
reviewed
Mar 31, 2026
Member
|
Please check our contribution guidelines (CONTRIBUTING.md). Your commit message should be no longer than 72 chars. Please wrap it! |
CMGS
force-pushed
the
fix/resume-clock-ordering
branch
from
March 31, 2026 06:39
c2324b3 to
477d212
Compare
Contributor
Author
|
@phip1611 thanks for your commnets, first time contribute to this hh, already wrap the commit logs and remove the empty line |
CMGS
force-pushed
the
fix/resume-clock-ordering
branch
from
March 31, 2026 06:48
477d212 to
4f39cdb
Compare
Member
|
We think we've seen this bug also in production - nice find! |
Reorder resume() to: set_clock, device_manager.resume, cpu_manager.resume. This matches the inverse of pause() which correctly saves the clock before pausing vCPUs. Signed-off-by: CMGS <[email protected]>
CMGS
force-pushed
the
fix/resume-clock-ordering
branch
from
March 31, 2026 06:57
4f39cdb to
62bac4c
Compare
rbradford
approved these changes
Mar 31, 2026
rbradford
left a comment
Member
There was a problem hiding this comment.
Cool, thanks! - i've also been investigating clock issues after a pause/resume. Not sure if this is the same issue but the fix lgtm!
rbradford
enabled auto-merge
March 31, 2026 07:06
phip1611
pushed a commit
to phip1611/cloud-hypervisor
that referenced
this pull request
Mar 31, 2026
Backport [0]. Reorder resume() to: set_clock, device_manager.resume, cpu_manager.resume. This matches the inverse of pause() which correctly saves the clock before pausing vCPUs. [0] cloud-hypervisor#7932 Signed-off-by: CMGS <[email protected]>
tpressure
pushed a commit
to cyberus-technology/cloud-hypervisor
that referenced
this pull request
Mar 31, 2026
Backport [0]. Reorder resume() to: set_clock, device_manager.resume, cpu_manager.resume. This matches the inverse of pause() which correctly saves the clock before pausing vCPUs. [0] cloud-hypervisor#7932 Signed-off-by: CMGS <[email protected]>
Coffeeri
added a commit
to Coffeeri/cloud-hypervisor
that referenced
this pull request
Apr 10, 2026
Since we backported the resume ordering from upstream PR cloud-hypervisor#7932 in cyberus PR cloud-hypervisor#133, virtio devices resume before vCPUs. Some devices need follow-up work that may notify the guest and must therefore wait until vCPUs are running again. We add a post-vCPU-resume virtio callback and invoke it after CpuManager::resume() completes. This callback will be used in a follow-up commit which drains the virtio-blk queue once the vCPUs are resumed after a restore. Ref: cloud-hypervisor#7932 Ref: cyberus-technology#133 On-behalf-of: SAP [email protected] Signed-off-by: Leander Kohler <[email protected]>
Coffeeri
added a commit
to cyberus-technology/cloud-hypervisor
that referenced
this pull request
Apr 10, 2026
Since we backported the resume ordering from upstream PR cloud-hypervisor#7932 in cyberus PR #133, virtio devices resume before vCPUs. Some devices need follow-up work that may notify the guest and must therefore wait until vCPUs are running again. We add a post-vCPU-resume virtio callback and invoke it after CpuManager::resume() completes. This callback will be used in a follow-up commit which drains the virtio-blk queue once the vCPUs are resumed after a restore. Ref: cloud-hypervisor#7932 Ref: #133 On-behalf-of: SAP [email protected] Signed-off-by: Leander Kohler <[email protected]>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The resume() method previously started vCPUs before calling KVM_SET_CLOCK, causing guests to execute with stale clock parameters. This is particularly problematic for Windows/Hyper-V guests where the TSC reference page contains outdated scale/offset values, leading to ~4 minute hangs after vm.restore.
Reorder resume() to: set_clock → device_manager.resume → cpu_manager.resume. This matches the inverse of pause() which correctly saves the clock before pausing vCPUs.
close #7930