LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
From: David Hildenbrand <email@example.com>
To: Peter Xu <firstname.lastname@example.org>,
Cc: Alistair Popple <email@example.com>,
Tiberiu Georgescu <firstname.lastname@example.org>,
Mike Rapoport <email@example.com>,
Hugh Dickins <firstname.lastname@example.org>,
Matthew Wilcox <email@example.com>,
Andrea Arcangeli <firstname.lastname@example.org>,
"Kirill A . Shutemov" <email@example.com>,
Andrew Morton <firstname.lastname@example.org>,
Mike Kravetz <email@example.com>
Subject: Re: [PATCH RFC 0/4] mm: Enable PM_SWAP for shmem with PTE_MARKER
Date: Tue, 17 Aug 2021 11:04:18 +0200 [thread overview]
Message-ID: <firstname.lastname@example.org> (raw)
Hi Peter, a couple of comments, sorry for the late reply.
> [Based on v5.14-rc4]
> This patchset enables PM_SWAP of pagemap on shmem. IOW userspace will be able
> to detect whether a shmem page is swapped out, just like anonymous pages.
> This feature can be enabled with CONFIG_PTE_MARKER_PAGEOUT. When enabled, it
> brings 0.8% overhead on swap-in performance on a shmem page, so I didn't make
> it the default yet. However IMHO 0.8% is still in an acceptable range that we
> can even make it the default at last. Comments are welcomed here.
Special config option and added complexity for handling a corner case
feature partially more correct. Hm.
> There's one previous series that wanted to address the same issue but in
> another way by Tiberiu A Georgescu <email@example.com>, here:
> In that series it's done by looking up page cache for all none ptes. However I
> raised concern on 4x performance degradation for all shmem pagemap users.
Who cares? I am asking because for me, it's hard to imagine a workload
that actually cares about a 4x performance degradation when querying the
pagemap in very special cases, especially if it involves gigantic shmem
ranges. VM live migration -- sync will be a bit slower? CRIU sync will
be a bit slower? I mean, actual page dumping is a lot more expensive.
Really a problem?
I read that CRIU cares about shmem via pagemap , at least for
anonymous shared memory; not sure how memfd is treated, I assume in a
similar way. But I do wonder how it even works reliably, because it
relies on present/swapped out and sofrtdirty tracking, which are both
essentially broken e.g., when swapping out AFAIKT. Looks like this
really needs a proper fix.
> Unlike the other approach, this series has zero overhead on pagemap read
> because the PM_SWAP info is consolidated into the zapped PTEs directly.
> One major goal of this series is to add the PM_SWAP support, the reason is as
> stated by Tiberiu and Ivan in the other patchset:
> As a summary: for some reason the userspace needs to scan the pages in the
> background, however that scanning could misguide page reclaim on which page is
> hot and which is cold. With correct PM_SWAP information, the userspace can
> correct the behavior of page reclaim by firstly fetching that info from
> pagemap, and explicit madvise(MADV_PAGEOUT). In this case, the pages are for
> the guest, but it can be any shmem page.
> Another major goal of this series is to do a proof-of-concept of the PTE marker
> idea, and that's also the major reason why it's RFC. So far PTE marker can
> potentially be the solution for below three problems that I'm aware of:
> (a) PM_SWAP on shmem
> (b) Userfaultfd-wp on shmem/hugetlbfs
> (c) PM_SOFT_DIRTY lost for shmem over swapping
> This series tries to resolve problem (a) which should be the simplest, ideally
> it should solve immediate problem for the live migration issue raised by
> Tiberiu and Ivan on proactive paging out unused guest pages.
> Both (a) and (c) will be for performance-wise or statistic-wise.
> Scenario (b) will require pte markers as part of the function to trap writes to
> uffd-wp protected regions when the pages were e.g. swapped out or zapped for
> any reason.
> Currently, uffd-wp shmem work (still during review on the list, latest v5, )
> used another solution called "special swap pte". It works similarly like PTE
> markers as both of the approachs are to persist information into zapped pte,
> but people showed concern about that idea and it's suggested to use a safer
> (swp-entry level operation, not pte level), and arch-independent approach.
> Hopefully PTE markers satifsfy these demands.
> Before I rework the uffd-wp series, I wanted to know whether this approach can
> be accepted upstream. So besides the swap part, comments on PTE markers will
> be extremely welcomed.
For uffd-wp in its current form, it would certainly be the way to go I
think. AFAIKT, the idea of special swap entries isn't new, just that
it's limited to anonymous memory for now, which makes things like fork
and new mappings a lot cheaper.
> What is PTE Markers?
> PTE markers are defined as some special PTEs that works like a "marker" just
> like in normal life. Firstly it uses a swap type, IOW it's not a valid/present
> pte, so processor will trigger a page fault when it's accessed. Meanwhile, the
> format of the PTE is well-defined, so as to contain some information that we
> would like to know before/during the page access happening.
> In this specific case, when the shmem page is paged out, we set a marker
> showing that this page was paged out, then when pagemap is read about this pte,
> we know this is a swapped-out/very-cold page.
> This use case is not an obvious one but the most simplest. The uffd-wp use
> case is more obvious (wr-protect is per-pte, so we can't save into page cache;
> meanwhile we need that info to persist across zappings e.g. thp split or page
> out of shmem pages).
> So in the future, it can contain more information, e.g., whether this pte is
> wr-protected by userfaultfd; whether this pte was written in this mm context
> for soft-dirtying. On 64 bit systems, we have a total of 58 bits (swp_offset).
> I'm also curious whether it can be further expanded to other mm areas. E.g.,
> logically it can work too for non-RAM based memories outside shmem/hugetlbfs,
> e.g. a common file system like ext4 or btrfs? As long as there will be a need
> to store some per-pte information across zapping of the ptes, then maybe it can
> be considered.
As already expressed, we should try storing as little information in
page tables as possible if we're dealing with shared memory. The
features we design around this all seem to over-complicate the actual
users, over-complicate fork, over-complicate handling on new mappings.
For uffd-wp in its current form, there seems to be no way around it, and
PTE markers seem to be what you want -- but as I already raised, the
feature itself on shmem is somewhat suboptimal, just like SOFTDIRTY
tracking on shmem.
But I guess I'm biased at this point because the main users of these
features actually want to query/set such properties for all sharers, not
individual processes; so the opinion of others would be appreciated.
> Known Issues/Concerns
> About THP
> Currently we don't need to worry about THP because paged out shmem pages will
> be split when shrinking, IOW we only need to consider PTE, and the markers will
> only be applied to a shmem pte not pmd or bigger.
> About PM_SWAP Accuracy
> This is not an "accurate" solution to provide PM_SWAP bit. Two exmaples:
> - When process A & B both map shmem page P somewhere, it can happen that only
> one of these ptes got marked with the pte marker. Imagine below sequence:
> 0. Process A & B both map shmem page P somewhere
> 1. Process A zap pte of page P for some reason (e.g. thp split)
> 2. System decides to recycle page P
> 3. System replace process B's pte (pointed to P) by PTE marker
> 4. System _didn't_ replace process A's pte because it was none pte, and
> it'll continue to be none pte
> 5. Only process B's relevant pte has the PTE marker after P swapped out
> - When fork, we don't copy shmem vma ptes, including the pte markers. So
> even if page P was swapped out, only the parent process has the pte marker
> installed, in child it'll be none pte if fork() happened after pageout.
> Conclusion: just like it used to be, the PM_SWAP is best-effort. But it should
> work in 99.99% cases and it should already start to solve problems.
At least I don't like these semantics at all. PM_SWAP is a cached value
which might be under-represented and consequently wrong. Take CRIU as an
example, it has to be correct even if a process would remap a memory
region, fork() and unmap in the parent as far as I understand, ...
If we really care about performance for users with the old semantics,
introduce some runtime toggle that enables the new behavior (even for a
single process?) and consequently is a bit slower in corner cases. But I
really do wonder if we care at all about the performance degradation in
If we really care about performance for users with new semantics, then
let's do it properly and see how we can actually speed it up without
per-process page table hacks.
Anyhow, just my 2 cents.
David / dhildenb
next prev parent reply other threads:[~2021-08-17 9:04 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-08-07 3:25 Peter Xu
2021-08-07 3:25 ` [PATCH RFC 1/4] mm: Introduce PTE_MARKER swap entry Peter Xu
2021-08-07 3:25 ` [PATCH RFC 2/4] mm: Check against orig_pte for finish_fault() Peter Xu
2021-08-07 3:25 ` [PATCH RFC 3/4] mm: Handle PTE_MARKER page faults Peter Xu
2021-08-07 3:25 ` [PATCH RFC 4/4] mm: Install marker pte when page out for shmem pages Peter Xu
2021-08-13 15:18 ` Tiberiu Georgescu
2021-08-13 16:01 ` Peter Xu
2021-08-18 18:02 ` Tiberiu Georgescu
2021-08-17 9:04 ` David Hildenbrand [this message]
2021-08-17 17:09 ` [PATCH RFC 0/4] mm: Enable PM_SWAP for shmem with PTE_MARKER Peter Xu
2021-08-17 18:46 ` David Hildenbrand
2021-08-17 20:24 ` Peter Xu
2021-08-18 8:24 ` David Hildenbrand
2021-08-18 17:52 ` Tiberiu Georgescu
2021-08-18 18:13 ` David Hildenbrand
2021-08-19 14:54 ` Tiberiu Georgescu
2021-08-19 17:26 ` David Hildenbrand
2021-08-20 16:49 ` Tiberiu Georgescu
2021-08-20 19:12 ` Peter Xu
2021-08-25 13:40 ` Tiberiu Georgescu
2021-08-25 14:59 ` Peter Xu
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:
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--subject='Re: [PATCH RFC 0/4] mm: Enable PM_SWAP for shmem with PTE_MARKER' \
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
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).