Fixes a race condition in killing Sandboxes #959
+1,153
−169
Conversation
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
simongdavies
requested review from
danbugs,
dblnz,
devigned,
jprendes,
ludfjig,
marosset and
syntactically
as code owners
simongdavies
force-pushed
the
fix-race-1
branch
3 times, most recently
from
96a6fce to
d200c7c
Compare
jsturtevant
reviewed
Oct 17, 2025
| /// retrying until either: | ||
| /// - The signal is successfully delivered (VCPU transitions from running to not running) | ||
| /// - The VCPU stops running for another reason (e.g., call completes normally) | ||
| /// - No call is active (call_active=false) |
There was a problem hiding this comment.
Does it block? I thought it would just return false?
jsturtevant
reviewed
Oct 18, 2025
jsturtevant
reviewed
Oct 18, 2025
| // The virtualization stack can use this function to return the control | ||
| // of a virtual processor back to the virtualization stack in case it | ||
| // needs to change the state of a VM or to inject an event into the processor | ||
| debug!("Internal cancellation detected, returning Retry error"); |
There was a problem hiding this comment.
simongdavies
force-pushed
the
fix-race-1
branch
from
4e211f2 to
4de305a
Compare
ludfjig
reviewed
Oct 21, 2025
| /// This is acceptable because the generation tracking provides an additional | ||
| /// safety layer. Even if a stale kill somehow stamped cancel_requested, the | ||
| /// generation mismatch would cause it to be ignored. | ||
| call_active: AtomicBool, |
There was a problem hiding this comment.
I think this flag is redundant. We will already not send any signals when calling kill if the vcpu is not running
Signed-off-by: Simon Davies <[email protected]>
Signed-off-by: James Sturtevant <[email protected]>
…sted by us Signed-off-by: James Sturtevant <[email protected]>
Signed-off-by: James Sturtevant <[email protected]>
Signed-off-by: James Sturtevant <[email protected]>
Signed-off-by: James Sturtevant <[email protected]>
Signed-off-by: James Sturtevant <[email protected]>
Signed-off-by: James Sturtevant <[email protected]>
Signed-off-by: James Sturtevant <[email protected]>
Signed-off-by: James Sturtevant <[email protected]>
Signed-off-by: Simon Davies <[email protected]>
Signed-off-by: Simon Davies <[email protected]>
Signed-off-by: Simon Davies <[email protected]>
Signed-off-by: Simon Davies <[email protected]>
Signed-off-by: Simon Davies <[email protected]>
Signed-off-by: Simon Davies <[email protected]>
Signed-off-by: Simon Davies <[email protected]>
Signed-off-by: Simon Davies <[email protected]>
Signed-off-by: Simon Davies <[email protected]>
Signed-off-by: Simon Davies <[email protected]>
Signed-off-by: Simon Davies <[email protected]>
Signed-off-by: Simon Davies <[email protected]>
Signed-off-by: Simon Davies <[email protected]>
simongdavies
force-pushed
the
fix-race-1
branch
from
cf14cfd to
f1aef3e
Compare
Signed-off-by: Simon Davies <[email protected]>
Signed-off-by: Simon Davies <[email protected]>
ludfjig
reviewed
Oct 23, 2025
| /// | ||
| /// This is only called on Linux (KVM/MSHV). Windows uses a different interrupt mechanism. | ||
| #[cfg(any(kvm, mshv))] | ||
| fn increment_call_generation(&self) -> u64; |
There was a problem hiding this comment.
these all method are public right now. I recommend making a new private trait InterruptHandleImpl like so pub(crate) trait InterruptHandleImpl: InterruptHandle and store a dyn of this trait, but return it as a dyn InterruptHandle to the user
ludfjig
reviewed
Oct 23, 2025
| } | ||
|
|
||
| // Default implementations for common methods | ||
| impl dyn InterruptHandle { |
There was a problem hiding this comment.
I believe you can put default method directly in the trait definition, no need to impl dyn InterruptHandle
ludfjig
reviewed
Oct 23, 2025
| ) -> Result<ReturnValue> { | ||
| // Mark that a guest function call is now active | ||
| // (This also increments the generation counter internally) | ||
| self.vm.interrupt_handle().set_call_active(); |
There was a problem hiding this comment.
You could consider making this a function which takes closure like
start_call(||{
...rest of existing code
})
where the function calls set_call_active and clear_call_active for you (and calls your closure in between).
Similar to existing maybe_time_and_emit_host_call
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.
Fixes a race condition where a sandbox kill arrives after a sandbox has successfully exited causing the subsequent run to fail.
There is a breaking change in this PR, previously if kill was called on an
InterruptHandlebefore or while a guest call was not in progress the next guest call made on theSandboxwould be cancelled , now this scenario is a no-op. kill only takes effect if there is a guest call running.