[DocDB] Use std::mutex for write lock in RWCLock #25391
Assignees
Labels
area/docdb
YugabyteDB core features
kind/enhancement
This is an enhancement of an existing feature
priority/medium
Medium priority issue
Comments
spolitov
added
area/docdb
yugabyte-ci
added
kind/enhancement
spolitov
added a commit
that referenced
this issue
Jan 10, 2025
Summary: Thread sanitizer does not understand `std::timed_mutex`, so it cannot detect related issue, such as lock order reversal. Could use `std::mutex` when building for TSAN to turn on detection of related issues. Also fixed all TSAN issues found after this change. There are two most common patterns for lock order reversion: 1) Usually we acquire object write lock while holding catalog manager mutex_. But sometimes we acquire catalog manager mutex_ while holding object write lock. Fixed by changing place where we acquire catalog manager mutex_. Doing it before acquiring object write lock. Or after object write lock is released. 2) Different write lock order for catalog objects. Fixed by adjusting code to take locks in the object id order. Jira: DB-14620 Test Plan: Jenkins Reviewers: zdrudi, hsunder, xCluster, asrivastava, stiwary Reviewed By: zdrudi, hsunder, asrivastava Subscribers: stiwary, yql, esheng, ybase Tags: #jenkins-ready Differential Revision: https://phorge.dev.yugabyte.com/D40697
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Jira Link: DB-14620
Description
Thread sanitizer does not understand timed_mutex, so it cannot detect related issue, such as lock order reversal.
Could use
std::mutexwhen building for TSAN to turn on detection of related issues.Issue Type
kind/enhancement
Warning: Please confirm that this issue does not contain any sensitive information
The text was updated successfully, but these errors were encountered: