bigquery/storage/managedwriter: Potential deadlock on error of ManagedStream in AppendRows

[jonomacd]

Client

BigQuery

Environment

local

Go Environment

any go env

Code

https://github.com/googleapis/google-cloud-go/blob/bigquery/v1.54.0/bigquery/storage/managedwriter/managed_stream.go#L293-L297

	ms.mu.Lock()
	if ms.err != nil {
		return nil, ms.err
	}
	ms.mu.Unlock()

Expected behavior

The managed writer should check an error and release the lock

Actual behavior

If an error case is detected on the ManagedStream we return early and will not release the mutex. Other writers using this stream will deadlock.

Suggested Fix

	ms.mu.Lock()
        err := ms.err
        ms.mu.Unlock()
	
       if err != nil {
		return nil, err
	}

EDIT: The same issue can be seen here in appendWithRetry
https://github.com/googleapis/google-cloud-go/blob/bigquery/v1.54.0/bigquery/storage/managedwriter/managed_stream.go#L199-L203

[shollyman]

Thanks for the report. Appears I introduced these while dealing with the context changes in #8275, will have a PR out shortly.