Fix slice backing-array race in live-store FindTraceByID#6968
Open
MukundaKatta wants to merge 4 commits intografana:mainfrom
Open
Fix slice backing-array race in live-store FindTraceByID#6968MukundaKatta wants to merge 4 commits intografana:mainfrom
MukundaKatta wants to merge 4 commits intografana:mainfrom
Conversation
MukundaKatta
requested review from
carles-grafana,
electron0zero,
ie-pham,
javiermolinar,
mapno,
mattdurham,
mdisibio,
oleg-kozlyuk-grafana,
ruslan-mikhailov,
stoewer,
yvrhdn and
zalegrala
as code owners
|
|
mattdurham
reviewed
Apr 15, 2026
|
|
||
| tr := &tempopb.Trace{ | ||
| ResourceSpans: liveTrace.Batches, | ||
| // Cap the slice to its length so appends below (and in any downstream |
Contributor
There was a problem hiding this comment.
Feels a bit verbose comment here, can we tone it down a bit.
Contributor
There was a problem hiding this comment.
Applies to other comments as well.
| @@ -1,5 +1,6 @@ | |||
| ## main / unreleased | |||
|
|
|||
| * [BUGFIX] Fix a panic in live-store `FindTraceByID` caused by a slice backing-array race between the trace combiner's append and concurrent proto.Marshal. The live trace's `Batches` slice is now handed to the combiner with its capacity capped to its length so downstream appends cannot mutate the backing array still referenced by the live trace. [#6958](https://github.com/grafana/tempo/issues/6958) | |||
Member
There was a problem hiding this comment.
changelog entry not in order.
Author
|
recheck |
FindTraceByID previously assigned the live trace's Batches slice header
directly to the temporary tempopb.Trace handed to the combiner:
tempTrace.ResourceSpans = liveTrace.Batches
Because the slice kept its original capacity, a later append by the
combiner (or by the push path that holds writeHeadBlock's tr.ResourceSpans)
would reuse the same backing array still referenced by liveTrace.Batches.
Concurrent proto.Marshal calls could then observe a slot being rewritten
between Size() and MarshalToSizedBuffer() and panic with a negative-index
`slice bounds out of range` error. The same pattern existed in
writeHeadBlock.
Cap the capacity to the length at both call sites so every consumer gets
its own backing array on append:
tempTrace.ResourceSpans = liveTrace.Batches[:len:len]
Add a regression test that pushes a trace, keeps it in liveTraces
(no cut), calls FindByTraceID, and asserts the returned ResourceSpans
slice has cap == len — the exact invariant the fix guarantees.
Fixes grafana#6958
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
MukundaKatta
force-pushed
the
fix/livestore-findtracebyid-slice-race
branch
from
0f94966 to
f2589bc
Compare
Contributor
|
Hello, are you able to sign the CLA? |
Author
|
Pushed 198a1ac addressing the review:
Also: on the CLA — I'll sign / chase the status so this can proceed. Thanks for the review! |
Align with the CHANGELOG convention used by the other entries in this section (PR link instead of issue link + trailing @author).
Author
|
@mattdurham @electron0zero — comment tone-down + CHANGELOG ordering were already addressed in 198a1ac (4-line comment in both |
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
4 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.
Summary
FindTraceByIDin the live store handed the combiner the live trace'sBatchesslice header directly. The slice kept its original capacity, so a later append by the combiner (or by the push path inwriteHeadBlock) reused the same backing array still referenced byliveTrace.Batches. Concurrentproto.Marshalcould then see a slot being rewritten betweenSize()andMarshalToSizedBuffer()and panic withslice bounds out of range [-N:].instance_search.goandinstance.go) so every consumer gets its own backing array on append.FindByTraceID, and asserts the returnedResourceSpansslice hascap == len— the invariant the fix guarantees.[BUGFIX]CHANGELOG entry.Fixes #6958. Re-submits the fix proposed in the now-closed #6964 (closed for CLA reasons) with the test and changelog the reviewer requested.
Test plan
TestInstanceFindByTraceIDLiveTraceSliceIsolationassertscap(resp.Trace.ResourceSpans) == len(resp.Trace.ResourceSpans)when the trace is served fromliveTraces.make test, lint, CLA)