Re: [PATCH 1/2] mm: kmsan: fix iounmap metadata teardown

[Dmytro Koziuk]

>  AI review might have a found a couple of issues - please check?

I don't think the patch needs to be changed for these issues. For
normal iounmap(), the
caller must ensure that the mapping is no longer in use. Also,
the x86 implementation keeps the virtual range reserved until after
kmsan_iounmap_page_range() returns and its TLB flush completes.

For the kmsan_ioremap_page_range() error path, the mapping has not yet
been returned to the caller. Cleanup and the TLB flush complete before
the reserved virtual range is released.


ср, 16 сент. 2026 г. в 01:35, Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>:
>
> On Tue, 15 Sep 2026 19:02:06 +0300 Dima Koziuk <dmytrokoziuk68@xxxxxxxxx> wrote:
>
> > While studying the code, I noticed that kmsan_iounmap_page_range() calls
> > __vunmap_range_noflush(v_shadow, vmalloc_shadow(end)) inside its per-page
> > loop, and does the same for origin. The first iteration therefore unmaps
> > the entire metadata range, removing the PTEs for later pages before the
> > loop can recover their backing pages.
>
> Thanks.
>
> AI review might have a found a couple of issues - please check?
>
> https://sashiko.dev/#/patchset/20260915160207.2952-1-dmytrokoziuk68@xxxxxxxxx