Re: [PATCH v4 2/5] mm: LARGE_ANON_FOLIO for improved performance

From: David Hildenbrand
Date: Fri Aug 04 2023 - 17:15:05 EST


On 04.08.23 23:00, Yu Zhao wrote:
On Fri, Aug 4, 2023 at 2:23 PM David Hildenbrand <david@xxxxxxxxxx> wrote:

On 04.08.23 10:27, Ryan Roberts wrote:
On 04/08/2023 00:50, Yu Zhao wrote:
On Thu, Aug 3, 2023 at 6:43 AM Ryan Roberts <ryan.roberts@xxxxxxx> wrote:

+ Kirill

On 26/07/2023 10:51, Ryan Roberts wrote:
Introduce LARGE_ANON_FOLIO feature, which allows anonymous memory to be
allocated in large folios of a determined order. All pages of the large
folio are pte-mapped during the same page fault, significantly reducing
the number of page faults. The number of per-page operations (e.g. ref
counting, rmap management lru list management) are also significantly
reduced since those ops now become per-folio.

The new behaviour is hidden behind the new LARGE_ANON_FOLIO Kconfig,
which defaults to disabled for now; The long term aim is for this to
defaut to enabled, but there are some risks around internal
fragmentation that need to be better understood first.

When enabled, the folio order is determined as such: For a vma, process
or system that has explicitly disabled THP, we continue to allocate
order-0. THP is most likely disabled to avoid any possible internal
fragmentation so we honour that request.

Otherwise, the return value of arch_wants_pte_order() is used. For vmas
that have not explicitly opted-in to use transparent hugepages (e.g.
where thp=madvise and the vma does not have MADV_HUGEPAGE), then
arch_wants_pte_order() is limited to 64K (or PAGE_SIZE, whichever is
bigger). This allows for a performance boost without requiring any
explicit opt-in from the workload while limitting internal
fragmentation.

If the preferred order can't be used (e.g. because the folio would
breach the bounds of the vma, or because ptes in the region are already
mapped) then we fall back to a suitable lower order; first
PAGE_ALLOC_COSTLY_ORDER, then order-0.


...

+#define ANON_FOLIO_MAX_ORDER_UNHINTED \
+ (ilog2(max_t(unsigned long, SZ_64K, PAGE_SIZE)) - PAGE_SHIFT)
+
+static int anon_folio_order(struct vm_area_struct *vma)
+{
+ int order;
+
+ /*
+ * If THP is explicitly disabled for either the vma, the process or the
+ * system, then this is very likely intended to limit internal
+ * fragmentation; in this case, don't attempt to allocate a large
+ * anonymous folio.
+ *
+ * Else, if the vma is eligible for thp, allocate a large folio of the
+ * size preferred by the arch. Or if the arch requested a very small
+ * size or didn't request a size, then use PAGE_ALLOC_COSTLY_ORDER,
+ * which still meets the arch's requirements but means we still take
+ * advantage of SW optimizations (e.g. fewer page faults).
+ *
+ * Finally if thp is enabled but the vma isn't eligible, take the
+ * arch-preferred size and limit it to ANON_FOLIO_MAX_ORDER_UNHINTED.
+ * This ensures workloads that have not explicitly opted-in take benefit
+ * while capping the potential for internal fragmentation.
+ */
+
+ if ((vma->vm_flags & VM_NOHUGEPAGE) ||
+ test_bit(MMF_DISABLE_THP, &vma->vm_mm->flags) ||
+ !hugepage_flags_enabled())
+ order = 0;
+ else {
+ order = max(arch_wants_pte_order(), PAGE_ALLOC_COSTLY_ORDER);
+
+ if (!hugepage_vma_check(vma, vma->vm_flags, false, true, true))
+ order = min(order, ANON_FOLIO_MAX_ORDER_UNHINTED);
+ }
+
+ return order;
+}


Hi All,

I'm writing up the conclusions that we arrived at during discussion in the THP
meeting yesterday, regarding linkage with exiting THP ABIs. It would be great if
I can get explicit "agree" or disagree + rationale from at least David, Yu and
Kirill.

In summary; I think we are converging on the approach that is already coded, but
I'd like confirmation.



The THP situation today
-----------------------

- At system level: THP can be set to "never", "madvise" or "always"
- At process level: THP can be "never" or "defer to system setting"
- At VMA level: no-hint, MADV_HUGEPAGE, MADV_NOHUGEPAGE

That gives us this table to describe how a page fault is handled, according to
process state (columns) and vma flags (rows):

| never | madvise | always
----------------|-----------|-----------|-----------
no hint | S | S | THP>S
MADV_HUGEPAGE | S | THP>S | THP>S
MADV_NOHUGEPAGE | S | S | S

Legend:
S allocate single page (PTE-mapped)
LAF allocate lage anon folio (PTE-mapped)
THP allocate THP-sized folio (PMD-mapped)
fallback (usually because vma size/alignment insufficient for folio)



Principles for Large Anon Folios (LAF)
--------------------------------------

David tells us there are use cases today (e.g. qemu live migration) which use
MADV_NOHUGEPAGE to mean "don't fill any PTEs that are not explicitly faulted"
and these use cases will break (i.e. functionally incorrect) if this request is
not honoured.

I don't remember David saying this. I think he was referring to UFFD,
not MADV_NOHUGEPAGE, when discussing what we need to absolutely
respect.

My understanding was that MADV_NOHUGEPAGE was being applied to regions *before*
UFFD was being registered, and the app relied on MADV_NOHUGEPAGE to not back any
unfaulted pages. It's not completely clear to me how not honouring
MADV_NOHUGEPAGE would break things though. David?

Sorry, I'm still lagging behind on some threads.

Imagine the following for VM postcopy live migration:

(1) Set MADV_NOHUGEPAGE on guest memory and discard all memory (e.g.,
MADV_DONTNEED), to start with a clean slate.
(2) Migrates some pages during precopy from the source and stores them
into guest memory on the destination. Some of the memory locations
will have pages populated.
(3) At some point, decide to enable postcopy: enable userfaultfd on
guest memory.
(4) Discard *selected* pages again that have been dirtied in the
meantime on the source. These are pages that have been migrated
previously.
(5) Start running the VM on the destination.
(6) Anything that's not populated will trigger userfaultfd missing
faults. Then, you can request them from the source and place them.

Assume you would populate more than required during 2), you can end up
not getting userfaultfd faults during 4) and corrupt your guest state.
It works if during (2) you migrated all guest memory, or if during 4)
you zap everything that still needs migr

I see what you mean now. Thanks.

Yes, in this case we have to interpret MADV_NOHUGEPAGE as nothing >4KB.

Note that it's still even unclear to me why we want to *not* call these things THP. It would certainly make everything less confusing if we call them THP, but with additional attributes.

I think that is one of the first things we should figure out because it also indirectly tells us what all these toggles mean and how/if we should redefine them (and if they even apply).

Currently THP == PMD size

In 2016, Hugh already envisioned PUD/PGD THP (see 49920d28781d ("mm: make transparent hugepage size public")) when he explicitly exposed "hpage_pmd_size". Not "hpage_size".

For hugetlb on arm64 we already support various sizes that are < PMD size and *not* call them differently. It's a huge(tlb) page. Sometimes we refer to them as cont-PTE hugetlb pages.


So, nowadays we do have "PMD-sized THP", someday we might have "PUD-sized THP". Can't we come up with a name to describe sub-PMD THP?

Is it really of value if we invent a new term for them? Yes, I was not enjoying "Flexible THP".


Once we figured that out, we should figure out if MADV_HUGEPAGE meant "only PMD-sized THP" or anything else?

Also, we can then figure out if MADV_NOHUGEPAGE meant "only PMD-sized THP" or anything else?


The simplest approach to me would be "they imply any THP, and once we need more tunables we might add some", similar to what Kirill also raised.


Again, it's all unclear to me at this point and I'm happy to hear opinions, because I really don't know.

--
Cheers,

David / dhildenb