GHSA-42cj-99w8-cp2p: Race Condition
Summary
go.opentelemetry.io/otel/bridge/opentracing introduced an unsynchronized extraBaggageItems map on bridgeSpan. One goroutine can write this map through Span.SetBaggageItem while another goroutine reads and iterates it during correlation baggage propagation, which can trigger Go's fatal concurrent map access panic and crash the process. The finding is low severity because exploitation requires a specific OpenTracing bridge configuration and concurrent use of the same span.
Introduced in commit: 8cddf30
Details
bridge/opentracing/bridge.go:80-85 adds extraBaggageItems map[string]string to bridgeSpan without a mutex or other synchronization primitive. bridge/opentracing/bridge.go:219-234 shows SetBaggageItem calling updateOtelContext, which lazily creates the map and writes s.extraBaggageItems[restrictedKey] = value without locking. bridge/opentracing/bridge.go:359-377 shows correlationGetHook reading bSpan.extraBaggageItems, checking len(items), and iterating for k, v := range items without locking. The finding evidence also identifies api/correlation/context.go:160-165 as the path where correlation.MapFromContext invokes the get hook, allowing a read path to run concurrently with baggage writes.
Because Go maps are not safe for concurrent read/write access, concurrent SetBaggageItem and correlation.MapFromContext calls on the same hooked bridgeSpan can terminate the process with a runtime error such as fatal error: concurrent map read and map write or fatal error: concurrent map iteration and map write.
PoC
validation-artifact.zip
The validation artifact contains a PoC at validation-artifact.tar:validationpocconcurrentmap.go and supporting notes at validation-artifact.tar:validationpocREADME.txt.
Use a checkout of pellared/opentelemetry-go at commit 8cddf30 with Go module downloads enabled. The local validation environment could not complete the run because GOPROXY=off blocked dependency resolution; that blocked output is saved in validation-artifact.tar:validationpocrun.log.
Commands:
sh cd /path/to/opentelemetry-go git checkout 8cddf30 tar -xOf /path/to/finding-directory/validation-artifact.tar validationpocconcurrentmap.go > ./validationpocconcurrentmap.go GOPROXY=https://proxy.golang.org,direct go run ./validationpocconcurrentmap.go
The PoC starts a BridgeTracer, creates a span, installs correlation hooks with tracer.NewHookedContext(ctx), initializes baggage once, then runs one goroutine repeatedly calling span.SetBaggageItem(...) while another repeatedly calls otelcorrelation.MapFromContext(ctx). A vulnerable build is expected to terminate with a Go runtime concurrent map access error, for example:
text fatal error: concurrent map read and map write
or:
text fatal error: concurrent map iteration and map write
Impact
This is a race condition / improper synchronization vulnerability in a shared Go map. Applications using the OpenTelemetry OpenTracing bridge with correlation hooks can crash if the same bridgeSpan is accessed concurrently, with one execution path setting baggage and another propagating correlation baggage. The practical impact is denial of service for the affected application process; exposure depends on whether application request handling or internal concurrency can trigger those operations on the same span.
Affected Software
Remediation
Recommended actions to resolve this vulnerability, in priority order.
- Upgrade
Upgrade
go/go.opentelemetry.io/otel/bridge/opentracingto a version that resolves this vulnerability.Fixed in 1.45.0
Event History
Frequently Asked Questions
Which deployments are exposed to this issue?
Exposure requires use of the OpenTracing bridge with the specific correlation baggage propagation configuration that invokes the bridge get hook. The issue also requires concurrent use of the same span.
What concurrent operations trigger the crash condition?
One goroutine must call Span.SetBaggageItem, which writes to the span's extraBaggageItems map, while another reads and iterates that map during correlation baggage propagation. This can cause Go to terminate the process with a fatal concurrent map access panic.
How can an affected code path be identified?
Review whether the application uses bridge/opentracing bridge spans, calls SetBaggageItem on them, and performs correlation baggage propagation through correlation.MapFromContext concurrently on the same span. The unsafe map accesses are in bridge/opentracing/bridge.go.