On Tue, Jul 12, 2022 at 11:43 AM Matthew Wilcox <willy@xxxxxxxxxxxxx> wrote:
>
> I'm going to leave discussing the permissions aspect to the experts in
> that realm, but from a practical point of view, why do we allow the dedupe
> ioctl to investigate arbitrary byte ranges? If you're going to dedupe,
> it has to be block aligned (both start and length).
I agree, although I think that's a separate issue.
I suspect it should just check for inode->i_blkbits alignment. I think
that's what DIO does, and it seems like a sane minimum.
Linus
