fix(snapshot/x86_64): make sure TSC_DEADLINE MSR is non-zero #4630
Conversation
This file contains 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
On x86_64, we observed that when restoring from a snapshot, one of the vCPUs had MSR_IA32_TSC_DEADLINE cleared and never received TSC interrupts until the MSR is updated externally (eg by setting the system time). We believe this happens because the TSC interrupt is lost during snapshot taking process: the MSR is cleared, but the interrupt is not delivered to the guest, so the guest does not rearm the timer. A visible effect of that is failure to connect to a restored VM via SSH. This commit introduces a workaround. If when taking a snapshot, we see a zero MSR_IA32_TSC_DEADLINE, we replace its value with the MSR_IA32_TSC value from the same vCPU to make sure that the vCPU will continue to receive TSC interrupts. (cherry picked from commit 94b37cb) Signed-off-by: Nikita Kalyazin <kalyazin@amazon.com>
The TSC_DEADLINE MSR value is volatile is it is getting updated by the guest kernel based on the current TSC value. (cherry picked from commit 4402c82) Signed-off-by: Nikita Kalyazin <kalyazin@amazon.com>
The TSC_DEADLINE MSR value is volatile is it is getting updated by the guest kernel based on the current TSC value. (cherry picked from commit cee34ab) Signed-off-by: Nikita Kalyazin <kalyazin@amazon.com>
kalyazin
force-pushed
the
fix_tsc_deadline_1.8
branch
from
ceea8bd
to
345218e
Compare
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## firecracker-v1.8 #4630 +/- ##
====================================================
+ Coverage 82.14% 82.16% +0.01%
====================================================
Files 255 255
Lines 31285 31307 +22
====================================================
+ Hits 25700 25722 +22
Misses 5585 5585
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. |
kalyazin
changed the title
kalyazin
added
the
Status: Awaiting review
ShadowCurse
approved these changes
May 31, 2024
kalyazin
changed the title
zulinx86
approved these changes
Jun 5, 2024
roypat
changed the title
|
We have observed no more failures of this type on main since merging #4618, so are confident to proceed with this PR, too. |
xmarcalx
approved these changes
Jun 11, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Changes
Backport from #4618 .
This change introduces a workaround. If when taking a snapshot, we see a zero MSR_IA32_TSC_DEADLINE, we replace its value with the MSR_IA32_TSC value from the same vCPU to make sure the vCPU will continue to receive TSC interrupts.
Reason
On x86_64, we observed that when restoring from a snapshot, one of the vCPUs had MSR_IA32_TSC_DEADLINE cleared and never received TSC interrupts until the MSR is updated externally (eg by setting the system time).
We believe this happens because the TSC interrupt is lost during snapshot taking process: the MSR is cleared, but the interrupt is not delivered to the guest, so the guest does not rearm the timer.
A visible effect of that is failure to connect to a restored VM via SSH, similar to https://buildkite.com/firecracker/firecracker-pr-nightly/builds/1403#018f83db-5395-4656-8d9c-83b6fcfcfd54/50-1994 .
License Acceptance
By submitting this pull request, I confirm that my contribution is made under
the terms of the Apache 2.0 license. For more information on following Developer
Certificate of Origin and signing off your commits, please check
CONTRIBUTING.md.PR Checklist
[ ] If a specific issue led to this PR, this PR closes the issue.PR.
[ ] API changes follow the Runbook for Firecracker API changes.CHANGELOG.md.[ ] NewTODOs link to an issue.contribution quality standards.
rust-vmm.