Re: [PATCH v6 01/23] mm: Introduce PTE_MARKER swap entry

From: Alistair Popple <apopple@nvidia.com>
To: Peter Xu <peterx@redhat.com>
Cc: <linux-mm@kvack.org>, <linux-kernel@vger.kernel.org>,
	Axel Rasmussen <axelrasmussen@google.com>,
	Nadav Amit <nadav.amit@gmail.com>,
	Mike Rapoport <rppt@linux.vnet.ibm.com>,
	Hugh Dickins <hughd@google.com>,
	Mike Kravetz <mike.kravetz@oracle.com>,
	"Kirill A . Shutemov" <kirill@shutemov.name>,
	Jerome Glisse <jglisse@redhat.com>,
	Matthew Wilcox <willy@infradead.org>,
	Andrew Morton <akpm@linux-foundation.org>,
	David Hildenbrand <david@redhat.com>,
	Andrea Arcangeli <aarcange@redhat.com>
Subject: Re: [PATCH v6 01/23] mm: Introduce PTE_MARKER swap entry
Date: Tue, 7 Dec 2021 13:12:23 +1100	[thread overview]
Message-ID: <2190897.LADUpQAO2C@nvdebian> (raw)
In-Reply-To: <Yam9EezLTANRA+Rf@xz-m1.local>

On Friday, 3 December 2021 5:45:37 PM AEDT Peter Xu wrote:
> On Fri, Dec 03, 2021 at 04:35:38PM +1100, Alistair Popple wrote:
> > > > > +static inline pte_marker pte_marker_get(swp_entry_t entry)
> > > > > +{
> > > > > +	return swp_offset(entry) & PTE_MARKER_MASK;
> > > > 
> > > > I'm not sure the PTE_MARKER_MASK adds much, especially as we only have one
> > > > user. I don't see a problem with open-coding these kind of checks (ie.
> > > 
> > > It's more or less a safety belt to make sure anything pte_marker_get() returned
> > > will be pte_marker defined bits only.
> > > 
> > > > swp_offset(entry) & PTE_MARKER_UFFD_WP) as you kind of end up doing that anyway.
> > > > Alternatively if you want helper functions I think it would be better to define
> > > > them for each marker. Eg: is_pte_marker_uffd_wp().
> > > 
> > > Yes we can have something like is_pte_marker_uffd_wp(), I didn't do that
> > > explicitly because I want us to be clear that pte_marker is a bitmask, so
> > > calling "is_*" will be slightly opaque - strictly speaking it should be
> > > "pte_marker_has_uffd_wp_bit()" if there will be more bits defined, but then the
> > > name of the helper will look a bit odd too.  Hence I just keep the only
> > > interface to fetch the whole marker and use "&" in the call sites to check.
> > 
> > Why does a caller need to care if it's a bitmask or not though? Isn't that an
> > implementation detail that could be left to the "is_*" functions? I must admit
> > I'm still working through the rest of this series though - is it because you
> > end up storing some kind of value in the upper bits of the PTE marker?
> 
> Nop. I'm just afraid the caller could overlook the fact that it's a bitmask,
> then there can be code like:
> 
>   if (is_pte_marker_uffd_wp(*ptep) && drop_uffd_wp)
>       pte_clear(ptep)
> 
> While we should only do:
> 
>   if (is_pte_marker_uffd_wp(*ptep) && drop_uffd_wp)
>       // remove uffd-wp bit in the pte_marker, keep the reset bitmask

I'm not sure how having the helper function prevents or changes this though? In
fact I just noticed this in patch 8:

                             if (uffd_wp_resolve &&
                                    (pte_marker_get(entry) & PTE_MARKER_UFFD_WP)) {
                                        pte_clear(vma->vm_mm, addr, pte);
                                        pages++;
                                }

And if I'm understanding your point correctly isn't that wrong because if there
were other users of pte markers they would inadvertently get cleared? Unless of
course I've missed something - I haven't looked at patch 8 yet for context. To
help with the above situation I think you would need a helper for clearing
ptes.

> I could be worrying too much, there's no real user of it.  If you prefer the
> helper a lot I can add it in the new version.  Thanks,

It's not a massive issue, but I do think either defining a helper or open
coding the bit check is clearer. I think we can worry about other users if/when
they appear.

 - Alistair