On Tue, Mar 8, 2022 at 9:26 AM Linus Torvalds
<torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
>
> Your actual patch looks pretty nasty, though. We avoid marking it
> accessed on purpose (to avoid atomicity issues wrt hw-dirty bits etc),
> but still, that patch makes me go "there has to be a better way".
That better way might be to make q() not use GUP at all.
It doesn't want the page - it just wants to make sure the fault is
handled. For the fault_in_readable() case we already just do the user
access.
The only reason we don't do that for fault_in_safe_writeable() is that
we don't have a good non-destructive user write.
Hmm. The futex code actually uses "fixup_user_fault()" for this case.
Maybe fault_in_safe_writeable() should do that?
Linus
