https://bugzilla.kernel.org/show_bug.cgi?id=217965 --- Comment #59 from Ojaswin Mujoo (ojaswin.mujoo@xxxxxxx) --- Hi Matthew, Thanks for testing out the patch and for sharing the results with and without CR_BEST_AVAIL_LEN commented. Would it be possible for you to add perf probes as mentioned in comment 36 and share the result. That'd be helpful in understanding why CR_BEST_AVAIL_LEN is getting stuck. Basically, I am interested in seeing how the scan is happening in CR_GOAL_LEN_FAST vs CR_BEST_AVAIL_LEN. [1] https://bugzilla.kernel.org/show_bug.cgi?id=217965#c36 Also, another thing I'd request you to try is, instead of commenting out the code can you run: $ echo "0" > /sys/fs/ext4/<dev>/mb_best_avail_max_trim_order and rerun the test. This disables most of the code but keeps some of the initial logic. Might help zero down if the issue is in the initial logic or the disabled code. Thanks again, ojaswin -- You may reply to this email to add a comment. You are receiving this mail because: You are watching the assignee of the bug.
