Re: [PATCH v2 20/26] userfaultfd: wp: support write protection for userfault vma range

[Mike Rapoport]

On Tue, Feb 26, 2019 at 03:20:28PM +0800, Peter Xu wrote:
> On Tue, Feb 26, 2019 at 08:43:47AM +0200, Mike Rapoport wrote:
> > On Tue, Feb 26, 2019 at 02:06:27PM +0800, Peter Xu wrote:
> > > On Mon, Feb 25, 2019 at 10:52:34PM +0200, Mike Rapoport wrote:
> > > > On Tue, Feb 12, 2019 at 10:56:26AM +0800, Peter Xu wrote:
> > > > > From: Shaohua Li <shli@xxxxxx>
> > > > > 
> > > > > Add API to enable/disable writeprotect a vma range. Unlike mprotect,
> > > > > this doesn't split/merge vmas.
> > > > > 
> > > > > Cc: Andrea Arcangeli <aarcange@xxxxxxxxxx>
> > > > > Cc: Rik van Riel <riel@xxxxxxxxxx>
> > > > > Cc: Kirill A. Shutemov <kirill@xxxxxxxxxxxxx>
> > > > > Cc: Mel Gorman <mgorman@xxxxxxx>
> > > > > Cc: Hugh Dickins <hughd@xxxxxxxxxx>
> > > > > Cc: Johannes Weiner <hannes@xxxxxxxxxxx>
> > > > > Signed-off-by: Shaohua Li <shli@xxxxxx>
> > > > > Signed-off-by: Andrea Arcangeli <aarcange@xxxxxxxxxx>
> > > > > [peterx:
> > > > >  - use the helper to find VMA;
> > > > >  - return -ENOENT if not found to match mcopy case;
> > > > >  - use the new MM_CP_UFFD_WP* flags for change_protection
> > > > >  - check against mmap_changing for failures]
> > > > > Signed-off-by: Peter Xu <peterx@xxxxxxxxxx>
> > > > > ---
> > > > >  include/linux/userfaultfd_k.h |  3 ++
> > > > >  mm/userfaultfd.c              | 54 +++++++++++++++++++++++++++++++++++
> > > > >  2 files changed, 57 insertions(+)
> > > > > 
> > > > > diff --git a/include/linux/userfaultfd_k.h b/include/linux/userfaultfd_k.h
> > > > > index 765ce884cec0..8f6e6ed544fb 100644
> > > > > --- a/include/linux/userfaultfd_k.h
> > > > > +++ b/include/linux/userfaultfd_k.h
> > > > > @@ -39,6 +39,9 @@ extern ssize_t mfill_zeropage(struct mm_struct *dst_mm,
> > > > >  			      unsigned long dst_start,
> > > > >  			      unsigned long len,
> > > > >  			      bool *mmap_changing);
> > > > > +extern int mwriteprotect_range(struct mm_struct *dst_mm,
> > > > > +			       unsigned long start, unsigned long len,
> > > > > +			       bool enable_wp, bool *mmap_changing);
> > > > > 
> > > > >  /* mm helpers */
> > > > >  static inline bool is_mergeable_vm_userfaultfd_ctx(struct vm_area_struct *vma,
> > > > > diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
> > > > > index fefa81c301b7..529d180bb4d7 100644
> > > > > --- a/mm/userfaultfd.c
> > > > > +++ b/mm/userfaultfd.c
> > > > > @@ -639,3 +639,57 @@ ssize_t mfill_zeropage(struct mm_struct *dst_mm, unsigned long start,
> > > > >  {
> > > > >  	return __mcopy_atomic(dst_mm, start, 0, len, true, mmap_changing, 0);
> > > > >  }
> > > > > +
> > > > > +int mwriteprotect_range(struct mm_struct *dst_mm, unsigned long start,
> > > > > +			unsigned long len, bool enable_wp, bool *mmap_changing)
> > > > > +{
> > > > > +	struct vm_area_struct *dst_vma;
> > > > > +	pgprot_t newprot;
> > > > > +	int err;
> > > > > +
> > > > > +	/*
> > > > > +	 * Sanitize the command parameters:
> > > > > +	 */
> > > > > +	BUG_ON(start & ~PAGE_MASK);
> > > > > +	BUG_ON(len & ~PAGE_MASK);
> > > > > +
> > > > > +	/* Does the address range wrap, or is the span zero-sized? */
> > > > > +	BUG_ON(start + len <= start);
> > > > 
> > > > I'd replace these BUG_ON()s with
> > > > 
> > > > 	if (WARN_ON())
> > > > 		 return -EINVAL;
> > > 
> > > I believe BUG_ON() is used because these parameters should have been
> > > checked in userfaultfd_writeprotect() already by the common
> > > validate_range() even before calling mwriteprotect_range().  So I'm
> > > fine with the WARN_ON() approach but I'd slightly prefer to simply
> > > keep the patch as is to keep Jerome's r-b if you won't disagree. :)
> > 
> > Right, userfaultfd_writeprotect() should check these parameters and if it
> > didn't it was a bug indeed. But still, it's not severe enough to crash the
> > kernel.
> > 
> > I hope Jerome wouldn't mind to keep his r-b with s/BUG_ON/WARN_ON ;-)
> > 
> > With this change you can also add 
> > 
> > Reviewed-by: Mike Rapoport <rppt@xxxxxxxxxxxxx>
> 
> Thanks!  Though before I change anything... please note that the
> BUG_ON()s are really what we've done in existing MISSING code.  One
> example is userfaultfd_copy() which did validate_range() first, then
> in __mcopy_atomic() we've used BUG_ON()s.  They make sense to me
> becauase userspace should never be able to trigger it.  And if we
> really want to change the BUG_ON()s in this patch, IMHO we probably
> want to change the other BUG_ON()s as well, then that can be a
> standalone patch or patchset to address another issue...

Yeah, we have quite a lot of them, so doing the replacement in a separate
patch makes perfect sense.
 
> (and if we really want to use WARN_ON, I would prefer WARN_ON_ONCE, or
>  directly return the errors to avoid DOS).

Agree.

> I'll see how you'd prefer to see how I should move on with this patch.

Let's keep this patch as is and make the replacement on top of the WP
series. Feel free to add r-b.
 
> Thanks,
> 
> -- 
> Peter Xu
> 

-- 
Sincerely yours,
Mike.