[IMPALA-4409] ImpalaServer::CancelInternal() can deadlock

• Type: Bug
• Status: Resolved
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: Impala 2.7.0, Impala 2.8.0
• Fix Version/s: Impala 2.8.0
• Component/s: Distributed Exec
• Labels: hang
• Target Version: Impala 2.8.0
• Assignee: Tim Armstrong
• Reporter: Tim Armstrong
• Created: 31/Oct/16 21:55
• Updated: 12/Jan/17 06:11
• Resolved: 03/Nov/16 20:15

Description

While looking through the code to understand another potential bug, I found out that CancelInternal() violates the lock ordering in ImpalaServer by acquiring SessionState::lock_ while holding QueryExecState::lock_. This could cause a hang if a number of threads were holding the right locks at the right time. I traced out the lock dependencies using a directed graph, which I've attached.