| Bug #108860 | InnoDB clone I/O errors from worker threads report incorrect errno | ||
|---|---|---|---|
| Submitted: | 24 Oct 2022 11:36 | Modified: | 25 Oct 2022 5:54 |
| Reporter: | Laurynas Biveinis (OCA) | Email Updates: | |
| Status: | Verified | Impact on me: | |
| Category: | MySQL Server: InnoDB storage engine | Severity: | S3 (Non-critical) |
| Version: | 8.0.31 | OS: | Any |
| Assigned to: | CPU Architecture: | Any | |
| Tags: | Clone | ||
[25 Oct 2022 5:54]
MySQL Verification Team
Hello Laurynas, Thank you for the report and feedback! regards, Umesh
Description: Suppose an I/O error happens during InnoDB clone in one of the worker threads, for example ER_ERROR_ON_READ. This thread will call Clone_Task_Manager::set_error, which will save the error by setting the m_saved_error and m_err_file_name fields. At some point later the clone connection thread will call Clone_Task_Manager::handler_error_other_task, which, for I/O errors, will call my_error as follows: my_error(m_saved_error, MYF(0), m_err_file_name.c_str(), errno, my_strerror(errbuf, sizeof(errbuf), errno)); Here m_saved_error and m_err_file_name will be the ones set by the previous set_error call, but note that errno is also reported. This errno value, being thread-local, will have no relation to any errno of the worker thread at the time of error occurrence, and will be misleading as a part of this diagnostics. How to repeat: Code analysis. Suggested fix: Introduce m_errno field, set it in set_error, report it in handle_error_other_task.