Re: [PATCH] mm: kmemleak: confirm suspected leaks with a second scan

[Breno Leitao]

On Thu, Jul 09, 2026 at 06:24:47PM +0100, Catalin Marinas wrote:
> The kmemleak marking phase is not atomic. While the object graph is
> traversed, the kernel can modify pointers, free objects or allocate new
> ones. If a reference to an object is moved from one location to another,
> kmemleak scanning may miss it. We have explicit annotations like
> kmemleak_transient_leak() but identifying and maintaining them is not
> trivial.
> 
> Given that such transient leaks are short-lived, rather than just
> reporting such objects as leaks, do another scan to confirm the
> suspected objects. If no new leaks are found during the first scan, skip
> the confirmation one.
> 
> Signed-off-by: Catalin Marinas <catalin.marinas@xxxxxxx>
> Assisted-by: Claude:claude-opus-4-8
> Cc: Breno Leitao <leitao@xxxxxxxxxx>
> Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>

Tested-by: Breno Leitao <leitao@xxxxxxxxxx>
Reviewed-by: Breno Leitao <leitao@xxxxxxxxxx>

PS: This is now conflicting with commit 946935b07cc94 ("mm/kmemleak:
skip the remaining scan phases when interrupted"), which has just showed
up in linux-next. So, a respin might be required.

> BTW, I'll be away from Saturday for three weeks. I may reply
> occasionally but won't be able to test anything.

Let me know if you want me to respin it based on linux-next.

BTW: Really thanks for doing it. I will backport it to Meta's kernel to
pressure-test it and keep an eye on false positives.
--breno