[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <Ywk9CKIJMX3z6WIq@xz-m1.local>
Date: Fri, 26 Aug 2022 17:37:12 -0400
From: Peter Xu <peterx@...hat.com>
To: David Hildenbrand <david@...hat.com>
Cc: Alistair Popple <apopple@...dia.com>, linux-mm@...ck.org,
akpm@...ux-foundation.org, Nadav Amit <nadav.amit@...il.com>,
huang ying <huang.ying.caritas@...il.com>,
LKML <linux-kernel@...r.kernel.org>,
"Sierra Guiza, Alejandro (Alex)" <alex.sierra@....com>,
Felix Kuehling <Felix.Kuehling@....com>,
Jason Gunthorpe <jgg@...dia.com>,
John Hubbard <jhubbard@...dia.com>,
Ralph Campbell <rcampbell@...dia.com>,
Matthew Wilcox <willy@...radead.org>,
Karol Herbst <kherbst@...hat.com>,
Lyude Paul <lyude@...hat.com>, Ben Skeggs <bskeggs@...hat.com>,
Logan Gunthorpe <logang@...tatee.com>, paulus@...abs.org,
linuxppc-dev@...ts.ozlabs.org, stable@...r.kernel.org,
Huang Ying <ying.huang@...el.com>
Subject: Re: [PATCH v3 2/3] mm/migrate_device.c: Copy pte dirty bit to page
On Fri, Aug 26, 2022 at 06:46:02PM +0200, David Hildenbrand wrote:
> On 26.08.22 17:55, Peter Xu wrote:
> > On Fri, Aug 26, 2022 at 04:47:22PM +0200, David Hildenbrand wrote:
> >>> To me anon exclusive only shows this mm exclusively owns this page. I
> >>> didn't quickly figure out why that requires different handling on tlb
> >>> flushs. Did I perhaps miss something?
> >>
> >> GUP-fast is the magic bit, we have to make sure that we won't see new
> >> GUP pins, thus the TLB flush.
> >>
> >> include/linux/mm.h:gup_must_unshare() contains documentation.
> >
> > Hmm.. Shouldn't ptep_get_and_clear() (e.g., xchg() on x86_64) already
> > guarantees that no other process/thread will see this pte anymore
> > afterwards?
>
> You could have a GUP-fast thread that just looked up the PTE and is
> going to pin the page afterwards, after the ptep_get_and_clear()
> returned. You'll have to wait until that thread finished.
IIUC the early tlb flush won't protect concurrent fast-gup from happening,
but I think it's safe because fast-gup will check pte after pinning, so
either:
(1) fast-gup runs before ptep_get_and_clear(), then
page_try_share_anon_rmap() will fail properly, or,
(2) fast-gup runs during or after ptep_get_and_clear(), then fast-gup
will see that either the pte is none or changed, then it'll fail the
fast-gup itself.
>
> Another user that relies on this interaction between GUP-fast and TLB
> flushing is for example mm/ksm.c:write_protect_page()
>
> There is a comment in there explaining the interaction a bit more detailed.
>
> Maybe we'll be able to handle this differently in the future (maybe once
> this turns out to be an actual performance problem). Unfortunately,
> mm->write_protect_seq isn't easily usable because we'd need have to make
> sure we're the exclusive writer.
>
>
> For now, it's not too complicated. For PTEs:
> * try_to_migrate_one() already uses ptep_clear_flush().
> * try_to_unmap_one() already conditionally used ptep_clear_flush().
> * migrate_vma_collect_pmd() was the one case that didn't use it already
> (and I wonder why it's different than try_to_migrate_one()).
I'm not sure whether I fully get the point, but here one major difference
is all the rest handles one page, so a tlb flush alongside with the pte
clear sounds reasonable. Even if so try_to_unmap_one() was modified to use
tlb batching, but then I see that anon exclusive made that batching
conditional. I also have question there on whether we can keep using the
tlb batching even with anon exclusive pages there.
In general, I still don't see how stall tlb could affect anon exclusive
pages on racing with fast-gup, because the only side effect of a stall tlb
is unwanted page update iiuc, the problem is fast-gup doesn't even use tlb,
afaict..
Thanks,
--
Peter Xu
Powered by blists - more mailing lists