tracer: Fix race in spanContext.setSamplingPriority #2271
Merged
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
This function read c.trace.priority without holding a lock, then
called functions that modify it. This causes the race detector to
detect the following race. This race bug is causing flaky tests in a
program that uses dd-trace-go. The race is triggered when two
goroutines both set the ext.ManualKeep on spans that share a root.
I think it might plausibly cause a real bug bug, where c.updated gets
set to true, even though it the priority was not changed.
My attempt to fix it is to move the logic to detect if the priority
was modified into the setSamplingPriority function itself. This might
not be the right fix since I don't really understand what this does,
but it seems to pass the tests both with and without race detection.
The race detector report without this change:
WARNING: DATA RACE
Write at 0x00c0019a0a88 by goroutine 2435:
gopkg.in/DataDog/dd-trace-go.v1/ddtrace/tracer.(*trace).setSamplingPriorityLocked()
/Users/evan.jones/dd/dd-trace-go/ddtrace/tracer/spancontext.go:327 +0xc4
gopkg.in/DataDog/dd-trace-go.v1/ddtrace/tracer.(*trace).push()
/Users/evan.jones/dd/dd-trace-go/ddtrace/tracer/spancontext.go:359 +0x23c
gopkg.in/DataDog/dd-trace-go.v1/ddtrace/tracer.newSpanContext()
/Users/evan.jones/dd/dd-trace-go/ddtrace/tracer/spancontext.go:143 +0x904
gopkg.in/DataDog/dd-trace-go.v1/ddtrace/tracer.(*tracer).StartSpan()
/Users/evan.jones/dd/dd-trace-go/ddtrace/tracer/tracer.go:516 +0xba0
gopkg.in/DataDog/dd-trace-go.v1/ddtrace/tracer.(*tracer).newChildSpan()
/Users/evan.jones/dd/dd-trace-go/ddtrace/tracer/tracer_test.go:52 +0x94
gopkg.in/DataDog/dd-trace-go.v1/ddtrace/tracer.TestTraceManualKeepRace.func1.1()
/Users/evan.jones/dd/dd-trace-go/ddtrace/tracer/span_test.go:441 +0x84
Previous read at 0x00c0019a0a88 by goroutine 2438:
gopkg.in/DataDog/dd-trace-go.v1/ddtrace/tracer.(*spanContext).setSamplingPriority()
/Users/evan.jones/dd/dd-trace-go/ddtrace/tracer/spancontext.go:185 +0x1bc
gopkg.in/DataDog/dd-trace-go.v1/ddtrace/tracer.(*span).setSamplingPriorityLocked()
/Users/evan.jones/dd/dd-trace-go/ddtrace/tracer/span.go:263 +0x8c
gopkg.in/DataDog/dd-trace-go.v1/ddtrace/tracer.(*span).setTagBool()
/Users/evan.jones/dd/dd-trace-go/ddtrace/tracer/span.go:388 +0xdc
gopkg.in/DataDog/dd-trace-go.v1/ddtrace/tracer.(*span).SetTag()
/Users/evan.jones/dd/dd-trace-go/ddtrace/tracer/span.go:127 +0x134
gopkg.in/DataDog/dd-trace-go.v1/ddtrace/tracer.TestTraceManualKeepRace.func1.1()
/Users/evan.jones/dd/dd-trace-go/ddtrace/tracer/span_test.go:442 +0xa8
darccio
approved these changes
Oct 23, 2023
Comment on lines
295
to
296
| t.mu.Lock() | ||
| defer t.mu.Unlock() |
There was a problem hiding this comment.
Reviewing this it seems to me that the original issue comes from checking c.trace.priority != nil && *c.trace.priority != float64(p) outside of the highlighted mutex.
In general, the change seems safe.
There was a problem hiding this comment.
That is correct: this fix ensures we only read t.priority while holding the lock, which required returning the boolean from this function.
evanj
commented
Oct 23, 2023
Comment on lines
295
to
296
| t.mu.Lock() | ||
| defer t.mu.Unlock() |
There was a problem hiding this comment.
That is correct: this fix ensures we only read t.priority while holding the lock, which required returning the boolean from this function.
|
This PR is stale because it has been open 20 days with no activity. Remove stale label or comment or this will be closed in 10 days. |
|
This was flagged stale but I believe this probably can be merged? @darccio or others, is there anything else I can do to help here? |
|
@evanj I think we can merge. Let me update the branch, run the tests and check everything goes fine. |
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.
What does this PR do?
This function read c.trace.priority without holding a lock, then called functions that modify it. This causes the race detector to detect the following race. This race bug is causing flaky tests in a program that uses dd-trace-go. The race is triggered when two goroutines both set the ext.ManualKeep on spans that share a root. I think it might plausibly cause a real bug bug, where c.updated gets set to true, even though it the priority was not changed.
My attempt to fix it is to move the logic to detect if the priority was modified into the setSamplingPriority function itself. This might not be the right fix since I don't really understand what this does, but it seems to pass the tests both with and without race detection.
The race detector report without this change:
Motivation
Fix flaky tests in programs that use dd-trace-go.
Reviewer's Checklist
For Datadog employees:
@DataDog/security-design-and-guidance.Unsure? Have a question? Request a review!