Details
-
Bug
-
Status: Closed (View Workflow)
-
Critical
-
Resolution: Fixed
-
10.9, 10.10, 10.11, 11.4
Description
mleich provided an rr replay trace where InnoDB would crash soon after a SET GLOBAL innodb_log_file_size=4194304 was completed. The reason appears to be that in log_t::resize_write() we fail to wrap around the log_sys.resize_lsn when the start offset (disregarding the length of the mini-transaction) exceeds the log_sys.resize_target (the requested size of the file). We would only perform the wrap-around when the start offset + the length would wrap around.
In the trace that I analyzed, we would catch a SIGSEGV on an attempt to write to log_sys.buf at an offset somewhere between 4 MiB and 8 MiB while the size of the memory mapped region is 4 MiB.
Attachments
Issue Links
- is caused by
-
MDEV-27812 Allow innodb_log_file_size to change without server restart
-
- Closed
-