MDEV-34453 Trying to read 16384 bytes at 70368744161280 outside the b…

[mariadb-DebarunBanerjee]

• The Jira issue number for this PR is: MDEV-34453

Description

The issue is caused by a race between buf_page_create_low getting the
page from buffer pool hash and buf_LRU_free_page evicting it from LRU.

The issue is introduced in 10.6 by MDEV-27058
commit aaef2e1d8c843d1e40b1ce0c5199c3abb3c5da28
MDEV-27058: Reduce the size of buf_block_t and buf_page_t

The solution is buffer fix the page before releasing buffer pool mutex
in buf_page_create_low when x_lock_try fails to acquire the page latch.

Release Notes

TBD

How can this PR be tested?

It is a hard to repeat the issue as it is. If we enforce the try_lock failed block in storage/innobase/buf/buf0buf.cc
buf_page_create_low() by commenting the try lock code, the issue repeats in release mode.
// const bool got= bpage->lock.x_lock_try();
// if (!got)

./mtr --mem repeat=50 stress.ddl_innodb

If the changes are not amenable to automated testing, please explain why not and carefully describe how to test manually.

Basing the PR against the correct MariaDB version

• This is a new feature or a refactoring, and the PR is based against the main branch.
• This is a bug fix, and the PR is based against the earliest maintained branch in which the bug can be reproduced.

PR quality check

• I checked the CODING_STANDARDS.md file and my PR conforms to this where appropriate.
• For any trivial modifications to the PR, I am ok with the reviewer making the changes themselves.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}