Re: [PATCH 3/3] hugetlb: allocate vma lock for all sharable vmas

[Miaohe Lin]

On 2022/10/5 9:17, Mike Kravetz wrote:
> The hugetlb vma lock was originally designed to synchronize pmd sharing.
> As such, it was only necessary to allocate the lock for vmas that were
> capable of pmd sharing.  Later in the development cycle, it was discovered
> that it could also be used to simplify fault/truncation races as described
> in [1].  However, a subsequent change to allocate the lock for all vmas
> that use the page cache was never made.  A fault/truncation race could
> leave pages in a file past i_size until the file is removed.
> 
> Remove the previous restriction and allocate lock for all VM_MAYSHARE
> vmas.  Warn in the unlikely event of allocation failure.
> 
> Fixes: "hugetlb: clean up code checking for fault/truncation races"
> Signed-off-by: Mike Kravetz <mike.kravetz@xxxxxxxxxx>

I believe this is a feasible solution. Thanks.

Reviewed-by: Miaohe Lin <linmiaohe@xxxxxxxxxx>

Thanks,
Miaohe Lin