Message-ID: <6moaipt4rmc62ijy2rtjbmzb5phgjpygxgqeic3bljydlwhxls@qqzuqbvs5gnh>
Date: Fri, 31 Jan 2025 11:20:33 +1100
From: Alistair Popple <apopple@...dia.com>
To: David Hildenbrand <david@...hat.com>, linux-kernel@...r.kernel.org, 
	linux-doc@...r.kernel.org, dri-devel@...ts.freedesktop.org, linux-mm@...ck.org, 
	nouveau@...ts.freedesktop.org, Andrew Morton <akpm@...ux-foundation.org>, 
	Jérôme Glisse <jglisse@...hat.com>, Jonathan Corbet <corbet@....net>, Alex Shi <alexs@...nel.org>, 
	Yanteng Si <si.yanteng@...ux.dev>, Karol Herbst <kherbst@...hat.com>, Lyude Paul <lyude@...hat.com>, 
	Danilo Krummrich <dakr@...nel.org>, David Airlie <airlied@...il.com>, 
	Simona Vetter <simona@...ll.ch>, "Liam R. Howlett" <Liam.Howlett@...cle.com>, 
	Lorenzo Stoakes <lorenzo.stoakes@...cle.com>, Vlastimil Babka <vbabka@...e.cz>, Jann Horn <jannh@...gle.com>, 
	Pasha Tatashin <pasha.tatashin@...een.com>, Peter Xu <peterx@...hat.com>, Jason Gunthorpe <jgg@...dia.com>
Subject: Re: [PATCH v1 4/4] mm/memory: document restore_exclusive_pte()

On Thu, Jan 30, 2025 at 11:43:25AM +0100, Simona Vetter wrote:
> On Thu, Jan 30, 2025 at 11:27:37AM +1100, Alistair Popple wrote:
> > On Wed, Jan 29, 2025 at 12:58:02PM +0100, David Hildenbrand wrote:
> > > Let's document how this function is to be used, and why the requirement
> > > for the folio lock might maybe be dropped in the future.
> > 
> > Sorry, only just catching up on your other thread. The folio lock was to ensure
> > the GPU got a chance to make forward progress by mapping the page. Without it
> > the CPU could immediately invalidate the entry before the GPU had a chance to
> > retry the fault.
> > 
> > Obviously performance wise having such thrashing is terrible, so should
> > really be avoided by userspace, but the lock at least allowed such programs
> > to complete.
> 
> Imo this is not a legit use-case. If userspace concurrently (instead of
> clearly alternating) uses the same 4k page for gpu atomics and on the cpu,
> it just gets to keep the fallout.
> 
> Plus there's no guarantee that we hold the folio_lock long enough for the
> gpu to actually complete the atomic, so this isn't even really helping
> with forward progress even if this somehow would be a legit usecase.

Yes, agree it's not a legit real world use case. In practice though it was
useful for testing this and other driver code by thrashing to generate a lot
device/cpu faults and invalidations. Obviously "just for testing" is not a great
justification though, so if it's causing problems we could get rid of it.

> But this is also why thp is potentially an issue, because if thp
> constantly creates pmd entries that potentially causes false sharing and
> we do have thrashing that shouldn't happen.

Yeah, I don't we should extend this to thps.

 - Alistair

> -Sima
> 
> > > Signed-off-by: David Hildenbrand <david@...hat.com>
> > > ---
> > >  mm/memory.c | 25 +++++++++++++++++++++++++
> > >  1 file changed, 25 insertions(+)
> > > 
> > > diff --git a/mm/memory.c b/mm/memory.c
> > > index 46956994aaff..caaae8df11a9 100644
> > > --- a/mm/memory.c
> > > +++ b/mm/memory.c
> > > @@ -718,6 +718,31 @@ struct folio *vm_normal_folio_pmd(struct vm_area_struct *vma,
> > >  }
> > >  #endif
> > >  
> > > +/**
> > > + * restore_exclusive_pte - Restore a device-exclusive entry
> > > + * @vma: VMA covering @address
> > > + * @folio: the mapped folio
> > > + * @page: the mapped folio page
> > > + * @address: the virtual address
> > > + * @ptep: PTE pointer into the locked page table mapping the folio page
> > > + * @orig_pte: PTE value at @ptep
> > > + *
> > > + * Restore a device-exclusive non-swap entry to an ordinary present PTE.
> > > + *
> > > + * The folio and the page table must be locked, and MMU notifiers must have
> > > + * been called to invalidate any (exclusive) device mappings. In case of
> > > + * fork(), MMU_NOTIFY_PROTECTION_PAGE is triggered, and in case of a page
> > > + * fault MMU_NOTIFY_EXCLUSIVE is triggered.
> > > + *
> > > + * Locking the folio makes sure that anybody who just converted the PTE to
> > > + * a device-private entry can map it into the device, before unlocking it; so
> > > + * the folio lock prevents concurrent conversion to device-exclusive.
> > 
> > I don't quite follow this - a concurrent conversion would already fail
> > because the GUP in make_device_exclusive_range() would most likely cause
> > an unexpected reference during the migration. And if a migration entry
> > has already been installed for the device private PTE conversion then
> > make_device_exclusive_range() will skip it as a non-present entry anyway.
> > 
> > However s/device-private/device-exclusive/ makes sense - the intent was to allow
> > the device to map it before a call to restore_exclusive_pte() (ie. a CPU fault)
> > could convert it back to a normal PTE.
> > 
> > > + * TODO: the folio lock does not protect against all cases of concurrent
> > > + * page table modifications (e.g., MADV_DONTNEED, mprotect), so device drivers
> > > + * must already use MMU notifiers to sync against any concurrent changes
> > 
> > Right. It's expected drivers are using MMU notifiers to keep page tables in
> > sync, same as for hmm_range_fault().
> > 
> > > + * Maybe the requirement for the folio lock can be dropped in the future.
> > > + */
> > >  static void restore_exclusive_pte(struct vm_area_struct *vma,
> > >  		struct folio *folio, struct page *page, unsigned long address,
> > >  		pte_t *ptep, pte_t orig_pte)
> > > -- 
> > > 2.48.1
> > > 
> 
> -- 
> Simona Vetter
> Software Engineer, Intel Corporation
> http://blog.ffwll.ch

Powered by blists - more mailing lists