From: James Houghton <jthoughton@google.com>
To: Peter Xu <peterx@redhat.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>,
Muchun Song <songmuchun@bytedance.com>,
David Hildenbrand <david@redhat.com>,
David Rientjes <rientjes@google.com>,
Axel Rasmussen <axelrasmussen@google.com>,
Mina Almasry <almasrymina@google.com>,
Jue Wang <juew@google.com>,
Manish Mishra <manish.mishra@nutanix.com>,
"Dr . David Alan Gilbert" <dgilbert@redhat.com>,
linux-mm@kvack.org, linux-kernel@vger.kernel.org
Subject: Re: [RFC PATCH 20/26] hugetlb: add support for high-granularity UFFDIO_CONTINUE
Date: Wed, 20 Jul 2022 13:58:06 -0700 [thread overview]
Message-ID: <CADrL8HX1YbDArmPsGUH+7b6dNxDa3Xo+fd5npGgz=RFehSw6iA@mail.gmail.com> (raw)
In-Reply-To: <YtGh46Jr0EGpqW7s@xz-m1.local>
On Fri, Jul 15, 2022 at 10:20 AM Peter Xu <peterx@redhat.com> wrote:
>
> On Fri, Jul 15, 2022 at 09:58:10AM -0700, James Houghton wrote:
> > On Fri, Jul 15, 2022 at 9:21 AM Peter Xu <peterx@redhat.com> wrote:
> > >
> > > On Fri, Jun 24, 2022 at 05:36:50PM +0000, James Houghton wrote:
> > > > The changes here are very similar to the changes made to
> > > > hugetlb_no_page, where we do a high-granularity page table walk and
> > > > do accounting slightly differently because we are mapping only a piece
> > > > of a page.
> > > >
> > > > Signed-off-by: James Houghton <jthoughton@google.com>
> > > > ---
> > > > fs/userfaultfd.c | 3 +++
> > > > include/linux/hugetlb.h | 6 +++--
> > > > mm/hugetlb.c | 54 +++++++++++++++++++++-----------------
> > > > mm/userfaultfd.c | 57 +++++++++++++++++++++++++++++++----------
> > > > 4 files changed, 82 insertions(+), 38 deletions(-)
> > > >
> > > > diff --git a/fs/userfaultfd.c b/fs/userfaultfd.c
> > > > index e943370107d0..77c1b8a7d0b9 100644
> > > > --- a/fs/userfaultfd.c
> > > > +++ b/fs/userfaultfd.c
> > > > @@ -245,6 +245,9 @@ static inline bool userfaultfd_huge_must_wait(struct userfaultfd_ctx *ctx,
> > > > if (!ptep)
> > > > goto out;
> > > >
> > > > + if (hugetlb_hgm_enabled(vma))
> > > > + goto out;
> > > > +
> > >
> > > This is weird. It means we'll never wait for sub-page mapping enabled
> > > vmas. Why?
> > >
> >
> > `ret` is true in this case, so we're actually *always* waiting.
>
> Aha! Then I think that's another problem, sorry. :) See Below.
>
> >
> > > Not to mention hugetlb_hgm_enabled() currently is simply VM_SHARED, so it
> > > means we'll stop waiting for all shared hugetlbfs uffd page faults..
> > >
> > > I'd expect in the in-house postcopy tests you should see vcpu threads
> > > spinning on the page faults until it's serviced.
> > >
> > > IMO we still need to properly wait when the pgtable doesn't have the
> > > faulted address covered. For sub-page mapping it'll probably need to walk
> > > into sub-page levels.
> >
> > Ok, SGTM. I'll do that for the next version. I'm not sure of the
> > consequences of returning `true` here when we should be returning
> > `false`.
>
> We've put ourselves onto the wait queue, if another concurrent
> UFFDIO_CONTINUE happened and pte is already installed, I think this thread
> could be waiting forever on the next schedule().
>
> The solution should be the same - walking the sub-page pgtable would work,
> afaict.
>
> [...]
>
> > > > @@ -6239,14 +6241,16 @@ int hugetlb_mcopy_atomic_pte(struct mm_struct *dst_mm,
> > > > * registered, we firstly wr-protect a none pte which has no page cache
> > > > * page backing it, then access the page.
> > > > */
> > > > - if (!huge_pte_none_mostly(huge_ptep_get(dst_pte)))
> > > > + if (!hugetlb_pte_none_mostly(dst_hpte))
> > > > goto out_release_unlock;
> > > >
> > > > - if (vm_shared) {
> > > > - page_dup_file_rmap(page, true);
> > > > - } else {
> > > > - ClearHPageRestoreReserve(page);
> > > > - hugepage_add_new_anon_rmap(page, dst_vma, dst_addr);
> > > > + if (new_mapping) {
> > >
> > > IIUC you wanted to avoid the mapcount accountings when it's the sub-page
> > > that was going to be mapped.
> > >
> > > Is it a must we get this only from the caller? Can we know we're doing
> > > sub-page mapping already here and make a decision with e.g. dst_hpte?
> > >
> > > It looks weird to me to pass this explicitly from the caller, especially
> > > that's when we don't really have the pgtable lock so I'm wondering about
> > > possible race conditions too on having stale new_mapping values.
> >
> > The only way to know what the correct value for `new_mapping` should
> > be is to know if we had to change the hstate-level P*D to non-none to
> > service this UFFDIO_CONTINUE request. I'll see if there is a nice way
> > to do that check in `hugetlb_mcopy_atomic_pte`.
> > Right now there is no
>
> Would "new_mapping = dest_hpte->shift != huge_page_shift(hstate)" work (or
> something alike)?
This works in the hugetlb_fault case, because in the hugetlb_fault
case, we install the largest PTE possible. If we are mapping a page
for the first time, we will use an hstate-sized PTE. But for
UFFDIO_CONTINUE, we may be installing a 4K PTE as the first PTE for
the whole hpage.
>
> > race, because we synchronize on the per-hpage mutex.
>
> Yeah not familiar with that mutex enough to tell, as long as that mutex
> guarantees no pgtable update (hmm, then why we need the pgtable lock
> here???) then it looks fine.
Let me take a closer look at this. I'll have a more detailed
explanation for the next version of the RFC.
>
> [...]
>
> > > > @@ -335,12 +337,16 @@ static __always_inline ssize_t __mcopy_atomic_hugetlb(struct mm_struct *dst_mm,
> > > > copied = 0;
> > > > page = NULL;
> > > > vma_hpagesize = vma_kernel_pagesize(dst_vma);
> > > > + if (use_hgm)
> > > > + vma_altpagesize = PAGE_SIZE;
> > >
> > > Do we need to check the "len" to know whether we should use sub-page
> > > mapping or original hpage size? E.g. any old UFFDIO_CONTINUE code will
> > > still want the old behavior I think.
> >
> > I think that's a fair point; however, if we enable HGM and the address
> > and len happen to be hstate-aligned
>
> The address can, but len (note! not "end" here) cannot?
They both (dst_start and len) need to be hpage-aligned, otherwise we
won't be able to install hstate-sized PTEs. Like if we're installing
4K at the beginning of a 1G hpage, we can't install a PUD, because we
only want to install that 4K.
>
> > , we basically do the same thing as
> > if HGM wasn't enabled. It could be a minor performance optimization to
> > do `vma_altpagesize=vma_hpagesize` in that case, but in terms of how
> > the page tables are set up, the end result would be the same.
>
> Thanks,
Thanks!
>
> --
> Peter Xu
>
next prev parent reply other threads:[~2022-07-20 20:58 UTC|newest]
Thread overview: 125+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-06-24 17:36 [RFC PATCH 00/26] hugetlb: Introduce HugeTLB high-granularity mapping James Houghton
2022-06-24 17:36 ` [RFC PATCH 01/26] hugetlb: make hstate accessor functions const James Houghton
2022-06-24 18:43 ` Mina Almasry
[not found] ` <e55f90f5-ba14-5d6e-8f8f-abf731b9095e@nutanix.com>
2022-06-27 12:09 ` manish.mishra
2022-06-28 17:08 ` James Houghton
2022-06-29 6:18 ` Muchun Song
2022-06-24 17:36 ` [RFC PATCH 02/26] hugetlb: sort hstates in hugetlb_init_hstates James Houghton
2022-06-24 18:51 ` Mina Almasry
2022-06-27 12:08 ` manish.mishra
2022-06-28 15:35 ` James Houghton
2022-06-27 18:42 ` Mike Kravetz
2022-06-28 15:40 ` James Houghton
2022-06-29 6:39 ` Muchun Song
2022-06-29 21:06 ` Mike Kravetz
2022-06-29 21:13 ` James Houghton
2022-06-24 17:36 ` [RFC PATCH 03/26] hugetlb: add make_huge_pte_with_shift James Houghton
2022-06-24 19:01 ` Mina Almasry
2022-06-27 12:13 ` manish.mishra
2022-06-24 17:36 ` [RFC PATCH 04/26] hugetlb: make huge_pte_lockptr take an explicit shift argument James Houghton
2022-06-27 12:26 ` manish.mishra
2022-06-27 20:51 ` Mike Kravetz
2022-06-28 15:29 ` James Houghton
2022-06-29 6:09 ` Muchun Song
2022-06-29 21:03 ` Mike Kravetz
2022-06-29 21:39 ` James Houghton
2022-06-29 22:24 ` Mike Kravetz
2022-06-30 9:35 ` Muchun Song
2022-06-30 16:23 ` James Houghton
2022-06-30 17:40 ` Mike Kravetz
2022-07-01 3:32 ` Muchun Song
2022-06-24 17:36 ` [RFC PATCH 05/26] hugetlb: add CONFIG_HUGETLB_HIGH_GRANULARITY_MAPPING James Houghton
2022-06-27 12:28 ` manish.mishra
2022-06-28 20:03 ` Mina Almasry
2022-06-24 17:36 ` [RFC PATCH 06/26] mm: make free_p?d_range functions public James Houghton
2022-06-27 12:31 ` manish.mishra
2022-06-28 20:35 ` Mike Kravetz
2022-07-12 20:52 ` James Houghton
2022-06-24 17:36 ` [RFC PATCH 07/26] hugetlb: add hugetlb_pte to track HugeTLB page table entries James Houghton
2022-06-27 12:47 ` manish.mishra
2022-06-29 16:28 ` James Houghton
2022-06-28 20:25 ` Mina Almasry
2022-06-29 16:42 ` James Houghton
2022-06-28 20:44 ` Mike Kravetz
2022-06-29 16:24 ` James Houghton
2022-07-11 23:32 ` Mike Kravetz
2022-07-12 9:42 ` Dr. David Alan Gilbert
2022-07-12 17:51 ` Mike Kravetz
2022-07-15 16:35 ` Peter Xu
2022-07-15 21:52 ` Axel Rasmussen
2022-07-15 23:03 ` Peter Xu
2022-09-08 17:38 ` Peter Xu
2022-09-08 17:54 ` James Houghton
2022-06-24 17:36 ` [RFC PATCH 08/26] hugetlb: add hugetlb_free_range to free PT structures James Houghton
2022-06-27 12:52 ` manish.mishra
2022-06-28 20:27 ` Mina Almasry
2022-06-24 17:36 ` [RFC PATCH 09/26] hugetlb: add hugetlb_hgm_enabled James Houghton
2022-06-27 12:55 ` manish.mishra
2022-06-28 20:33 ` Mina Almasry
2022-09-08 18:07 ` Peter Xu
2022-09-08 18:13 ` James Houghton
2022-06-24 17:36 ` [RFC PATCH 10/26] hugetlb: add for_each_hgm_shift James Houghton
2022-06-27 13:01 ` manish.mishra
2022-06-28 21:58 ` Mina Almasry
2022-07-07 21:39 ` Mike Kravetz
2022-07-08 15:52 ` James Houghton
2022-07-09 21:55 ` Mina Almasry
2022-06-24 17:36 ` [RFC PATCH 11/26] hugetlb: add hugetlb_walk_to to do PT walks James Houghton
2022-06-27 13:07 ` manish.mishra
2022-07-07 23:03 ` Mike Kravetz
2022-09-08 18:20 ` Peter Xu
2022-06-24 17:36 ` [RFC PATCH 12/26] hugetlb: add HugeTLB splitting functionality James Houghton
2022-06-27 13:50 ` manish.mishra
2022-06-29 16:10 ` James Houghton
2022-06-29 14:33 ` manish.mishra
2022-06-29 16:20 ` James Houghton
2022-06-24 17:36 ` [RFC PATCH 13/26] hugetlb: add huge_pte_alloc_high_granularity James Houghton
2022-06-29 14:11 ` manish.mishra
2022-06-24 17:36 ` [RFC PATCH 14/26] hugetlb: add HGM support for hugetlb_fault and hugetlb_no_page James Houghton
2022-06-29 14:40 ` manish.mishra
2022-06-29 15:56 ` James Houghton
2022-06-24 17:36 ` [RFC PATCH 15/26] hugetlb: make unmapping compatible with high-granularity mappings James Houghton
2022-07-19 10:19 ` manish.mishra
2022-07-19 15:58 ` James Houghton
2022-06-24 17:36 ` [RFC PATCH 16/26] hugetlb: make hugetlb_change_protection compatible with HGM James Houghton
2022-06-24 17:36 ` [RFC PATCH 17/26] hugetlb: update follow_hugetlb_page to support HGM James Houghton
2022-07-19 10:48 ` manish.mishra
2022-07-19 16:19 ` James Houghton
2022-06-24 17:36 ` [RFC PATCH 18/26] hugetlb: use struct hugetlb_pte for walk_hugetlb_range James Houghton
2022-06-24 17:36 ` [RFC PATCH 19/26] hugetlb: add HGM support for copy_hugetlb_page_range James Houghton
2022-07-11 23:41 ` Mike Kravetz
2022-07-12 17:19 ` James Houghton
2022-07-12 18:06 ` Mike Kravetz
2022-07-15 21:39 ` Axel Rasmussen
2022-06-24 17:36 ` [RFC PATCH 20/26] hugetlb: add support for high-granularity UFFDIO_CONTINUE James Houghton
2022-07-15 16:21 ` Peter Xu
2022-07-15 16:58 ` James Houghton
2022-07-15 17:20 ` Peter Xu
2022-07-20 20:58 ` James Houghton [this message]
2022-07-21 19:09 ` Peter Xu
2022-07-21 19:44 ` James Houghton
2022-07-21 19:53 ` Peter Xu
2022-06-24 17:36 ` [RFC PATCH 21/26] hugetlb: add hugetlb_collapse James Houghton
2022-06-24 17:36 ` [RFC PATCH 22/26] madvise: add uapi for HugeTLB HGM collapse: MADV_COLLAPSE James Houghton
2022-06-24 17:36 ` [RFC PATCH 23/26] userfaultfd: add UFFD_FEATURE_MINOR_HUGETLBFS_HGM James Houghton
2022-06-24 17:36 ` [RFC PATCH 24/26] arm64/hugetlb: add support for high-granularity mappings James Houghton
2022-06-24 17:36 ` [RFC PATCH 25/26] selftests: add HugeTLB HGM to userfaultfd selftest James Houghton
2022-06-24 17:36 ` [RFC PATCH 26/26] selftests: add HugeTLB HGM to KVM demand paging selftest James Houghton
2022-06-24 18:29 ` [RFC PATCH 00/26] hugetlb: Introduce HugeTLB high-granularity mapping Matthew Wilcox
2022-06-27 16:36 ` James Houghton
2022-06-27 17:56 ` Dr. David Alan Gilbert
2022-06-27 20:31 ` James Houghton
2022-06-28 0:04 ` Nadav Amit
2022-06-30 19:21 ` Peter Xu
2022-07-01 5:54 ` Nadav Amit
2022-06-28 8:20 ` Dr. David Alan Gilbert
2022-06-30 16:09 ` Peter Xu
2022-06-24 18:41 ` Mina Almasry
2022-06-27 16:27 ` James Houghton
2022-06-28 14:17 ` Muchun Song
2022-06-28 17:26 ` Mina Almasry
2022-06-28 17:56 ` Dr. David Alan Gilbert
2022-06-29 18:31 ` James Houghton
2022-06-29 20:39 ` Axel Rasmussen
2022-06-24 18:47 ` Matthew Wilcox
2022-06-27 16:48 ` James Houghton
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='CADrL8HX1YbDArmPsGUH+7b6dNxDa3Xo+fd5npGgz=RFehSw6iA@mail.gmail.com' \
--to=jthoughton@google.com \
--cc=almasrymina@google.com \
--cc=axelrasmussen@google.com \
--cc=david@redhat.com \
--cc=dgilbert@redhat.com \
--cc=juew@google.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=manish.mishra@nutanix.com \
--cc=mike.kravetz@oracle.com \
--cc=peterx@redhat.com \
--cc=rientjes@google.com \
--cc=songmuchun@bytedance.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).