Re: [PATCH 0/2] mm: kmemleak: avoid long RCU read-side sections in reports
From: Weilin Tong
Date: Wed Nov 11 2026 - 02:45:04 EST
Hi Andrew,
Thanks again for the review. Following up on my earlier reply, I'll
withdraw patch 2 for now while I work on a simpler solution.
Counting visited objects alone still leaves an unbounded RCU section
when walking removed objects. I've tested alternatives, but they add
either substantial read overhead or more cursor and reference-management
complexity than I'd like. I need more time to find a better balance.
Patch 1 fixes an independent cleanup race and has Catalin's Reviewed-by,
so it can be considered separately.
Thanks,
Weilin