From: Jason Gunthorpe <jgg@ziepe.ca>
To: John Hubbard <jhubbard@nvidia.com>
Cc: Peter Xu <peterx@redhat.com>,
Linus Torvalds <torvalds@linux-foundation.org>,
Leon Romanovsky <leonro@nvidia.com>,
Linux-MM <linux-mm@kvack.org>,
Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
"Maya B . Gokhale" <gokhale2@llnl.gov>,
Yang Shi <yang.shi@linux.alibaba.com>,
Marty Mcfadden <mcfadden8@llnl.gov>,
Kirill Shutemov <kirill@shutemov.name>,
Oleg Nesterov <oleg@redhat.com>, Jann Horn <jannh@google.com>,
Jan Kara <jack@suse.cz>, Kirill Tkhai <ktkhai@virtuozzo.com>,
Andrea Arcangeli <aarcange@redhat.com>,
Christoph Hellwig <hch@lst.de>,
Andrew Morton <akpm@linux-foundation.org>
Subject: Re: [PATCH 1/4] mm: Trial do_wp_page() simplification
Date: Wed, 16 Sep 2020 14:48:04 -0300 [thread overview]
Message-ID: <20200916174804.GC8409@ziepe.ca> (raw)
In-Reply-To: <e6c352f8-7ee9-0702-10a4-122d2c4422fc@nvidia.com>
On Tue, Sep 15, 2020 at 06:50:46PM -0700, John Hubbard wrote:
> >
> > It seems very strange that a physical page exclusively owned by a
> > process can become copied if pin_user_pages() is active and the
> > process did fork() at some point.
> >
> > Could the new pin_user_pages() logic help here? eg the
> > GUP_PIN_COUNTING_BIAS stuff?
> >
> > Could the COW code consider a refcount of GUP_PIN_COUNTING_BIAS + 1 as
> > being owned by the current mm and not needing COW? The DMA pin would
> > be 'invisible' for COW purposes?
>
>
> Please do be careful to use the API, rather than the implementation. The
> FOLL_PIN refcounting system results in being able to get a "maybe
> DMA-pinned", or a "definitely not DMA-pinned", via this API call:
So, what I'm thinking is something like (untested):
diff --git a/mm/memory.c b/mm/memory.c
index 469af373ae76e1..77f63183667e52 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -2889,6 +2889,26 @@ static vm_fault_t wp_page_shared(struct vm_fault *vmf)
return ret;
}
+static bool cow_needed(struct vm_fault *vmf)
+{
+ int total_map_swapcount;
+
+ if (!reuse_swap_page(vmf->page, &total_map_swapcount)) {
+ unlock_page(vmf->page);
+ return true;
+ }
+
+ if (total_map_swapcount == 1) {
+ /*
+ * The page is all ours. Move it to our anon_vma so the rmap
+ * code will not search our parent or siblings. Protected
+ * against the rmap code by the page lock.
+ */
+ page_move_anon_rmap(vmf->page, vmf->vma);
+ }
+ return false;
+}
+
/*
* This routine handles present pages, when users try to write
* to a shared page. It is done by copying the page to a new address
@@ -2947,8 +2967,21 @@ static vm_fault_t do_wp_page(struct vm_fault *vmf)
if (!trylock_page(page))
goto copy;
if (PageKsm(page) || page_mapcount(page) != 1 || page_count(page) != 1) {
+ bool cow = true;
+
+ /*
+ * If the page is DMA pinned we can't rely on the
+ * above to know if there are other CPU references as
+ * page_count() will be elevated by the
+ * pin. Needlessly copying the page will cause the DMA
+ * pin to break, try harder to avoid that.
+ */
+ if (page_maybe_dma_pinned(page))
+ cow = cow_needed(vmf);
+
unlock_page(page);
- goto copy;
+ if (cow)
+ goto copy;
}
/*
* Ok, we've got the only map reference, and the only
What do you think Peter? Is this remotely close?
Seems like it preserves the fast path in most cases, the page_count &
page_maybe_dma_pinned could be further optimized down to one atomic in
non huge page cases.
Jason
next prev parent reply other threads:[~2020-09-16 19:02 UTC|newest]
Thread overview: 98+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-08-21 23:49 [PATCH 0/4] mm: Simplfy cow handling Peter Xu
2020-08-21 23:49 ` [PATCH 1/4] mm: Trial do_wp_page() simplification Peter Xu
2020-08-24 8:36 ` Kirill Tkhai
2020-08-24 14:30 ` Jan Kara
2020-08-24 15:37 ` Kirill Tkhai
2020-08-24 18:22 ` Linus Torvalds
2020-09-01 7:01 ` Hugh Dickins
2020-09-14 14:38 ` Jason Gunthorpe
2020-09-14 17:32 ` Linus Torvalds
2020-09-14 18:34 ` Peter Xu
2020-09-14 21:15 ` Peter Xu
2020-09-14 22:55 ` Jason Gunthorpe
2020-09-14 22:59 ` Linus Torvalds
2020-09-14 23:28 ` Jason Gunthorpe
2020-09-15 0:19 ` Linus Torvalds
2020-09-15 14:50 ` Peter Xu
2020-09-15 15:17 ` Peter Xu
2020-09-15 16:05 ` Jason Gunthorpe
2020-09-15 18:29 ` Jason Gunthorpe
2020-09-15 19:13 ` Peter Xu
2020-09-15 19:38 ` Jason Gunthorpe
2020-09-15 21:33 ` Peter Xu
2020-09-15 23:22 ` Jason Gunthorpe
2020-09-16 1:50 ` John Hubbard
2020-09-16 17:48 ` Jason Gunthorpe [this message]
2020-09-16 18:46 ` Peter Xu
2020-09-17 11:25 ` Jason Gunthorpe
2020-09-17 18:11 ` Linus Torvalds
2020-09-17 19:38 ` Jason Gunthorpe
2020-09-17 19:51 ` Linus Torvalds
2020-09-18 16:40 ` Peter Xu
2020-09-18 17:16 ` Linus Torvalds
2020-09-18 19:57 ` Peter Xu
2020-09-18 17:32 ` Jason Gunthorpe
2020-09-18 20:40 ` Peter Xu
2020-09-18 20:59 ` Linus Torvalds
2020-09-19 0:28 ` Jason Gunthorpe
2020-09-18 21:06 ` John Hubbard
2020-09-19 0:01 ` Jason Gunthorpe
2020-09-21 8:35 ` Jan Kara
2020-09-21 12:03 ` Jason Gunthorpe
2022-02-16 16:59 ` Oded Gabbay
2022-02-16 17:24 ` Oded Gabbay
2022-02-16 19:04 ` Linus Torvalds
2022-02-16 19:20 ` Oded Gabbay
2022-02-16 19:24 ` David Hildenbrand
2020-09-21 13:42 ` Michal Hocko
2020-09-21 14:18 ` Peter Xu
2020-09-21 14:28 ` Michal Hocko
2020-09-21 14:38 ` Tejun Heo
2020-09-21 14:43 ` Christian Brauner
2020-09-21 14:55 ` Michal Hocko
2020-09-21 15:04 ` Christian Brauner
2020-09-21 16:06 ` Michal Hocko
2020-09-23 7:53 ` Michal Hocko
2020-09-21 14:41 ` Christian Brauner
2020-09-21 14:57 ` Michal Hocko
2020-09-21 16:31 ` Peter Xu
2020-09-17 18:14 ` Peter Xu
2020-09-17 18:26 ` Linus Torvalds
2020-09-17 19:03 ` Peter Xu
2020-09-17 19:42 ` Linus Torvalds
2020-09-17 19:55 ` John Hubbard
2020-09-17 20:06 ` Jason Gunthorpe
2020-09-17 20:19 ` John Hubbard
2020-09-17 20:25 ` Jason Gunthorpe
2020-09-17 20:35 ` Linus Torvalds
2020-09-17 21:40 ` Peter Xu
2020-09-17 22:09 ` Jason Gunthorpe
2020-09-17 22:25 ` Linus Torvalds
2020-09-17 22:48 ` Ira Weiny
2020-09-18 9:36 ` Jan Kara
2020-09-18 9:44 ` Jan Kara
2020-09-18 16:19 ` Jason Gunthorpe
2020-09-15 10:23 ` Leon Romanovsky
2020-09-15 15:56 ` Jason Gunthorpe
2020-09-15 15:03 ` Oleg Nesterov
2020-09-15 16:18 ` Peter Xu
2020-08-21 23:49 ` [PATCH 2/4] mm/ksm: Remove reuse_ksm_page() Peter Xu
2020-08-21 23:49 ` [PATCH 3/4] mm/gup: Remove enfornced COW mechanism Peter Xu
2020-09-14 14:27 ` Oleg Nesterov
2020-09-14 17:59 ` Peter Xu
2020-09-14 19:03 ` Linus Torvalds
2020-08-21 23:49 ` [PATCH 4/4] mm: Add PGREUSE counter Peter Xu
2020-08-22 16:14 ` Linus Torvalds
2020-08-24 0:24 ` Peter Xu
2020-08-22 16:05 ` [PATCH 0/4] mm: Simplfy cow handling Linus Torvalds
2020-08-23 23:58 ` Peter Xu
2020-08-24 8:38 ` Kirill Tkhai
2020-08-27 14:15 ` Peter Xu
2021-02-02 14:40 [PATCH 1/4] mm: Trial do_wp_page() simplification Gal Pressman
2021-02-02 16:31 ` Peter Xu
2021-02-02 16:44 ` Jason Gunthorpe
2021-02-02 17:05 ` Peter Xu
2021-02-02 17:13 ` Jason Gunthorpe
2021-02-03 12:43 ` Gal Pressman
2021-02-03 14:00 ` Jason Gunthorpe
2021-02-03 14:47 ` Gal Pressman
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20200916174804.GC8409@ziepe.ca \
--to=jgg@ziepe.ca \
--cc=aarcange@redhat.com \
--cc=akpm@linux-foundation.org \
--cc=gokhale2@llnl.gov \
--cc=hch@lst.de \
--cc=jack@suse.cz \
--cc=jannh@google.com \
--cc=jhubbard@nvidia.com \
--cc=kirill@shutemov.name \
--cc=ktkhai@virtuozzo.com \
--cc=leonro@nvidia.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mcfadden8@llnl.gov \
--cc=oleg@redhat.com \
--cc=peterx@redhat.com \
--cc=torvalds@linux-foundation.org \
--cc=yang.shi@linux.alibaba.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).