Linux Trace Kernel
 help / color / mirror / Atom feed
From: "David Hildenbrand (Arm)" <david@kernel.org>
To: Nico Pache <npache@redhat.com>
Cc: linux-doc@vger.kernel.org, linux-kernel@vger.kernel.org,
	linux-mm@kvack.org, linux-trace-kernel@vger.kernel.org,
	aarcange@redhat.com, akpm@linux-foundation.org,
	anshuman.khandual@arm.com, apopple@nvidia.com, baohua@kernel.org,
	baolin.wang@linux.alibaba.com, byungchul@sk.com,
	catalin.marinas@arm.com, cl@gentwo.org, corbet@lwn.net,
	dave.hansen@linux.intel.com, dev.jain@arm.com, gourry@gourry.net,
	hannes@cmpxchg.org, hughd@google.com, jack@suse.cz,
	jackmanb@google.com, jannh@google.com, jglisse@google.com,
	joshua.hahnjy@gmail.com, kas@kernel.org, lance.yang@linux.dev,
	liam@infradead.org, ljs@kernel.org,
	mathieu.desnoyers@efficios.com, matthew.brost@intel.com,
	mhiramat@kernel.org, mhocko@suse.com, peterx@redhat.com,
	pfalcato@suse.de, rakie.kim@sk.com, raquini@redhat.com,
	rdunlap@infradead.org, richard.weiyang@gmail.com,
	rientjes@google.com, rostedt@goodmis.org, rppt@kernel.org,
	ryan.roberts@arm.com, shivankg@amd.com, sunnanyong@huawei.com,
	surenb@google.com, thomas.hellstrom@linux.intel.com,
	tiwai@suse.de, vbabka@suse.cz, vishal.moola@gmail.com,
	wangkefeng.wang@huawei.com, will@kernel.org, willy@infradead.org,
	yang@os.amperecomputing.com, ying.huang@linux.alibaba.com,
	ziy@nvidia.com, zokeefe@google.com,
	Usama Arif <usama.arif@linux.dev>,
	usamaarif642@gmail.com
Subject: Re: [PATCH mm-unstable v18 11/14] mm/khugepaged: Introduce mTHP collapse support
Date: Wed, 3 Jun 2026 11:55:31 +0200	[thread overview]
Message-ID: <b5616c79-8123-4717-b56e-b5be131eadef@kernel.org> (raw)
In-Reply-To: <19639b08-5bf1-4974-9635-c458d512fa38@redhat.com>

On 6/2/26 19:23, Nico Pache wrote:
> 
> 
> On 6/1/26 7:15 AM, David Hildenbrand (Arm) wrote:
>>>
>>> So I looked into your items below. It seems logical, and I think it
>>> works the same way; however, your method seems slightly harder to
>>> understand due to all the edge cases and more error-prone to future
>>> changes (the stack holds implicit knowledge of the offset/order that
>>> must now be tracked in the edge cases).
>>>
>>> Given the stack is 24 bytes, I'm not sure if the extra complexity is
>>> worth saving that small amount of memory. Although we would also be
>>> getting rid of (3?) functions, so both approaches have pros and cons.
>>
>> I consider a simple forward loop over the offset ... less complexity compared to
>> a stack structure :)
>>
>>>
>>> I will implement a patch comparing your solution against mine and send
>>> it here, then we can decide which approach is better.
>>
>> Right, throw it over the fence and I'll see how to improve it further.
> 
> Ok heres what the diff looks like on top of my V19. 
> 
> you can access the tree here https://gitlab.com/npache/linux/-/commits/mthp-v19?ref_type=heads for easier review.
> 
> So far I have no problem with this approach it appeared cleaner than i thought. Did some light testing. Gonna throw it more through the ringer tomorrow. 

It's very clean.

Almost too nice to be true ;)

[...]

>  	unsigned int nr_occupied_ptes, nr_ptes, max_ptes_none;
>  	enum scan_result last_result = SCAN_FAIL;
> -	int collapsed = 0, stack_size = 0;
> +	int collapsed = 0;
>  	bool alloc_failed = false;
>  	unsigned long collapse_address;
> -	struct mthp_range range;
> -	u16 offset;
> -	u8 order;
> +	unsigned int offset = 0;
> +	unsigned int order = HPAGE_PMD_ORDER;


In include/linux/huge_mm.h we have

	highest_order()

and

	next_order()

They essentially allow you to get rid of the test_bit() and just jump to the
next enabled order right away.

I assume with only a handful of enabled_orders, that might be much more efficient.

I tried to optimize it and ended with the following, which is completely untested.

I think it might make sense to defer that and start with the simple approach you have.

I do wonder, though, about the last hunk below: should we bail out early if
enabled_orders is suddenly 0?



From 0d8ff955b3071f354b7fc9b627820fa374fa99dc Mon Sep 17 00:00:00 2001
From: "David Hildenbrand (Arm)" <david@kernel.org>
Date: Wed, 3 Jun 2026 11:52:44 +0200
Subject: [PATCH] tmp

Signed-off-by: David Hildenbrand (Arm) <david@kernel.org>
---
 include/linux/huge_mm.h |   5 ++
 mm/khugepaged.c         | 132 ++++++++++++++++++++++------------------
 2 files changed, 78 insertions(+), 59 deletions(-)

diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
index 48496f09909b..099318bc1181 100644
--- a/include/linux/huge_mm.h
+++ b/include/linux/huge_mm.h
@@ -205,6 +205,11 @@ static inline int highest_order(unsigned long orders)
 	return fls_long(orders) - 1;
 }
 
+static inline int smallest_order(unsigned long orders)
+{
+	return __ffs(orders);
+}
+
 static inline int next_order(unsigned long *orders, int prev)
 {
 	*orders &= ~BIT(prev);
diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 6de935e76ceb..49be9d1a88cb 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -99,8 +99,6 @@ static DEFINE_READ_MOSTLY_HASHTABLE(mm_slots_hash, MM_SLOTS_HASH_BITS);
 
 static struct kmem_cache *mm_slot_cache __ro_after_init;
 
-#define KHUGEPAGED_MIN_MTHP_ORDER	2
-
 struct collapse_control {
 	bool is_khugepaged;
 
@@ -1454,76 +1452,86 @@ static enum scan_result collapse_huge_page(struct mm_struct *mm, unsigned long s
  */
 static enum scan_result mthp_collapse(struct mm_struct *mm,
 		unsigned long address, int referenced, int unmapped,
-		struct collapse_control *cc, unsigned long enabled_orders)
+		struct collapse_control *cc, const unsigned long enabled_orders)
 {
-	unsigned int nr_occupied_ptes, nr_ptes, max_ptes_none;
 	enum scan_result last_result = SCAN_FAIL;
 	int collapsed = 0;
 	bool alloc_failed = false;
 	unsigned long collapse_address;
 	unsigned int offset = 0;
-	unsigned int order = HPAGE_PMD_ORDER;
 
+	/* We cannot collapse anon folios to order-1 or order-0. */
+	VM_WARN_ON_ONCE(!enabled_order || (enabled_orders & 0x3));
 
 	while (offset < HPAGE_PMD_NR) {
-		nr_ptes = 1UL << order;
-
-		if (!test_bit(order, &enabled_orders))
-			goto next_order;
-
-		max_ptes_none = collapse_max_ptes_none(cc, NULL, order);
-		nr_occupied_ptes = bitmap_weight_from(cc->mthp_present_ptes, offset,
-						      offset + nr_ptes);
-
-		if (nr_occupied_ptes >= nr_ptes - max_ptes_none) {
-			enum scan_result ret;
-
-			collapse_address = address + offset * PAGE_SIZE;
-			ret = collapse_huge_page(mm, collapse_address, referenced,
-						 unmapped, cc, order);
-
-			switch (ret) {
-			/* Cases where we continue to next collapse candidate */
-			case SCAN_SUCCEED:
-				collapsed += nr_ptes;
-				fallthrough;
-			case SCAN_PTE_MAPPED_HUGEPAGE:
-				goto next_offset;
-			/* Cases where lower orders might still succeed */
-			case SCAN_ALLOC_HUGE_PAGE_FAIL:
-				alloc_failed = true;
-				fallthrough;
-			case SCAN_LACK_REFERENCED_PAGE:
-			case SCAN_EXCEED_NONE_PTE:
-			case SCAN_EXCEED_SWAP_PTE:
-			case SCAN_EXCEED_SHARED_PTE:
-			case SCAN_PAGE_LOCK:
-			case SCAN_PAGE_COUNT:
-			case SCAN_PAGE_NULL:
-			case SCAN_DEL_PAGE_LRU:
-			case SCAN_PTE_NON_PRESENT:
-			case SCAN_PTE_UFFD_WP:
-			case SCAN_PAGE_LAZYFREE:
-				last_result = ret;
-				goto next_order;
-			/* Cases where no further collapse is possible */
-			case SCAN_PMD_MAPPED:
-				fallthrough;
-			default:
-				last_result = ret;
-				goto done;
+		/*
+		 * We can only collapse to a maximum order for a given offset.
+		 * So ignore all orders that do not apply to the current
+		 * offset, then see if any order to collapse to remains.
+		 */
+		unsigned long orders = enabled_orders & GENMASK(__ffs(offset), 0);
+		unsigned int order = highest_order(orders);
+
+		while (order) {
+			const unsigned int nr_ptes = 1UL << order;
+			unsigned int nr_occupied_ptes, max_ptes_none;
+
+			max_ptes_none = collapse_max_ptes_none(cc, NULL, order);
+			nr_occupied_ptes = bitmap_weight_from(cc->mthp_present_ptes, offset,
+							      offset + nr_ptes);
+
+			if (nr_occupied_ptes >= nr_ptes - max_ptes_none) {
+				enum scan_result ret;
+
+				collapse_address = address + offset * PAGE_SIZE;
+				ret = collapse_huge_page(mm, collapse_address, referenced,
+							 unmapped, cc, order);
+
+				switch (ret) {
+				/* Cases where we continue to next collapse candidate */
+				case SCAN_SUCCEED:
+					collapsed += nr_ptes;
+					fallthrough;
+				case SCAN_PTE_MAPPED_HUGEPAGE:
+					goto next_offset;
+				/* Cases where lower orders might still succeed */
+				case SCAN_ALLOC_HUGE_PAGE_FAIL:
+					alloc_failed = true;
+					fallthrough;
+				case SCAN_LACK_REFERENCED_PAGE:
+				case SCAN_EXCEED_NONE_PTE:
+				case SCAN_EXCEED_SWAP_PTE:
+				case SCAN_EXCEED_SHARED_PTE:
+				case SCAN_PAGE_LOCK:
+				case SCAN_PAGE_COUNT:
+				case SCAN_PAGE_NULL:
+				case SCAN_DEL_PAGE_LRU:
+				case SCAN_PTE_NON_PRESENT:
+				case SCAN_PTE_UFFD_WP:
+				case SCAN_PAGE_LAZYFREE:
+					last_result = ret;
+					break;
+				/* Cases where no further collapse is possible */
+				case SCAN_PMD_MAPPED:
+					fallthrough;
+				default:
+					last_result = ret;
+					goto done;
+				}
 			}
-		}
 
-next_order:
-		if (order > KHUGEPAGED_MIN_MTHP_ORDER &&
-			(BIT(order) - 1) & enabled_orders) {
-			order = order - 1;
-			continue;
+			order = next_order(&orders, order);
 		}
+
 next_offset:
-		offset += nr_ptes;
-		order = min_t(int, __ffs(offset), HPAGE_PMD_ORDER);
+		/*
+		 * Continue with the next collapse candidate. If we do not
+		 * have an order, skip to nest smallest mTHP we can collapse to.
+		 */
+		if (order)
+			offset += 1UL << order;
+		else
+			offset = ALIGN(offset + 1, smallest_order(enabled_orders));
 	}
 done:
 	if (collapsed)
@@ -1567,6 +1575,12 @@ static enum scan_result collapse_scan_pmd(struct mm_struct *mm,
 
 	enabled_orders = collapse_allowable_orders(vma, vma->vm_flags, tva_flags);
 
+	if (unlikely(!enabled_orders)) {
+		cc->progress++;
+		result = SCAN_SUCCEED;
+		goto out;
+	}
+
 	/*
 	 * If PMD is the only enabled order, enforce max_ptes_none, otherwise
 	 * scan all pages to populate the bitmap for mTHP collapse.
-- 
2.43.0


-- 
Cheers,

David

  parent reply	other threads:[~2026-06-03  9:55 UTC|newest]

Thread overview: 91+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-05-22 14:59 [PATCH mm-hotfixes-unstable v18 00/14] khugepaged: add mTHP collapse support Nico Pache
2026-05-22 14:59 ` [PATCH mm-unstable v18 01/14] mm/khugepaged: generalize hugepage_vma_revalidate for mTHP support Nico Pache
2026-05-22 14:59 ` [PATCH mm-unstable v18 02/14] mm/khugepaged: generalize alloc_charge_folio() Nico Pache
2026-05-22 14:59 ` [PATCH mm-unstable v18 03/14] mm/khugepaged: rework max_ptes_* handling with helper functions Nico Pache
2026-05-22 21:16   ` David Hildenbrand (Arm)
2026-06-01 13:26   ` Lorenzo Stoakes
2026-05-22 14:59 ` [PATCH mm-unstable v18 04/14] mm/khugepaged: generalize __collapse_huge_page_* for mTHP support Nico Pache
2026-05-22 21:24   ` David Hildenbrand (Arm)
2026-05-26 14:39   ` Nico Pache
2026-06-01 14:04   ` Lorenzo Stoakes
2026-05-22 15:00 ` [PATCH mm-unstable v18 05/14] mm/khugepaged: require collapse_huge_page to enter/exit with the lock dropped Nico Pache
2026-06-01 14:07   ` Lorenzo Stoakes
2026-06-02 10:26     ` Nico Pache
2026-05-22 15:00 ` [PATCH mm-unstable v18 06/14] mm/khugepaged: generalize collapse_huge_page for mTHP collapse Nico Pache
2026-05-22 21:47   ` David Hildenbrand (Arm)
2026-05-26 14:42   ` Nico Pache
2026-05-31  9:39   ` Lance Yang
2026-05-31 20:00     ` David Hildenbrand (Arm)
2026-06-01  3:28       ` Lance Yang
2026-06-01  6:54         ` David Hildenbrand (Arm)
2026-06-01  7:49           ` Lance Yang
2026-06-01  8:15             ` David Hildenbrand (Arm)
2026-06-01  8:44               ` Lance Yang
2026-06-01 10:09                 ` David Hildenbrand (Arm)
2026-06-01  9:08           ` Lance Yang
2026-06-01 10:23             ` David Hildenbrand (Arm)
2026-06-01 10:47               ` Lance Yang
2026-06-01 11:13                 ` David Hildenbrand (Arm)
2026-06-01 15:00                   ` Nico Pache
2026-06-01 15:05                     ` David Hildenbrand (Arm)
2026-06-01 16:07                       ` Lance Yang
2026-06-02 15:30                 ` Nico Pache
2026-06-02 16:34                   ` Lance Yang
2026-05-22 15:00 ` [PATCH mm-unstable v18 07/14] mm/khugepaged: skip collapsing mTHP to smaller orders Nico Pache
2026-05-22 21:51   ` David Hildenbrand (Arm)
2026-05-22 15:00 ` [PATCH mm-unstable v18 08/14] mm/khugepaged: add per-order mTHP collapse failure statistics Nico Pache
2026-05-31 20:09   ` David Hildenbrand (Arm)
2026-06-01 14:13   ` Lorenzo Stoakes
2026-05-22 15:00 ` [PATCH mm-unstable v18 09/14] mm/khugepaged: improve tracepoints for mTHP orders Nico Pache
2026-05-22 15:00 ` [PATCH mm-unstable v18 10/14] mm/khugepaged: introduce collapse_allowable_orders helper function Nico Pache
2026-05-31 20:18   ` David Hildenbrand (Arm)
2026-06-01 14:35     ` Lorenzo Stoakes
2026-06-01 14:40       ` David Hildenbrand (Arm)
2026-05-22 15:00 ` [PATCH mm-unstable v18 11/14] mm/khugepaged: Introduce mTHP collapse support Nico Pache
2026-05-25 14:15   ` Nico Pache
2026-05-25 19:10     ` Andrew Morton
2026-05-26  6:57       ` Wei Yang
2026-05-26 12:07         ` Nico Pache
2026-05-28  8:42           ` Wei Yang
2026-05-28 17:11             ` Nico Pache
2026-05-31  7:18   ` Lance Yang
2026-05-31  8:48     ` Lance Yang
2026-06-01 12:01       ` Nico Pache
2026-06-01 12:06         ` David Hildenbrand (Arm)
2026-06-02 10:58     ` Nico Pache
2026-06-02 15:44       ` Lance Yang
2026-06-03  8:05         ` David Hildenbrand (Arm)
2026-06-01  8:11   ` David Hildenbrand (Arm)
2026-06-01 12:40     ` Nico Pache
2026-06-01 13:15       ` David Hildenbrand (Arm)
2026-06-02 17:23         ` Nico Pache
2026-06-02 17:26           ` Nico Pache
2026-06-03  9:55           ` David Hildenbrand (Arm) [this message]
2026-06-03 10:00           ` David Hildenbrand (Arm)
2026-06-03 12:16             ` Nico Pache
2026-06-03 12:27               ` David Hildenbrand (Arm)
2026-05-22 15:00 ` [PATCH mm-unstable v18 12/14] mm/khugepaged: avoid unnecessary mTHP collapse attempts Nico Pache
2026-05-31  7:31   ` Lance Yang
2026-05-31 20:02     ` David Hildenbrand (Arm)
2026-06-01  1:53       ` Lance Yang
2026-05-22 15:00 ` [PATCH mm-unstable v18 13/14] mm/khugepaged: run khugepaged for all orders Nico Pache
2026-05-22 15:00 ` [PATCH mm-unstable v18 14/14] Documentation: mm: update the admin guide for mTHP collapse Nico Pache
2026-05-22 21:58   ` David Hildenbrand (Arm)
2026-05-26 12:00     ` Nico Pache
2026-05-26 14:45   ` Nico Pache
2026-05-22 15:07 ` [PATCH mm-hotfixes-unstable v18 00/14] khugepaged: add mTHP collapse support Nico Pache
2026-05-22 15:13   ` Vlastimil Babka (SUSE)
2026-05-22 16:11     ` Nico Pache
2026-05-22 21:13       ` David Hildenbrand (Arm)
2026-05-22 15:16   ` Lorenzo Stoakes
2026-05-22 16:08     ` Nico Pache
2026-05-22 16:19       ` Lorenzo Stoakes
2026-05-22 16:31         ` Nico Pache
2026-05-22 17:12           ` Lorenzo Stoakes
2026-05-26  8:14             ` Lorenzo Stoakes
2026-05-22 15:13 ` Lorenzo Stoakes
2026-05-22 20:47 ` Andrew Morton
2026-06-01 15:58   ` Alexander Gordeev
2026-06-01 17:05     ` Nico Pache
2026-06-01 17:08     ` Lorenzo Stoakes
2026-06-02  1:53       ` Lance Yang

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=b5616c79-8123-4717-b56e-b5be131eadef@kernel.org \
    --to=david@kernel.org \
    --cc=aarcange@redhat.com \
    --cc=akpm@linux-foundation.org \
    --cc=anshuman.khandual@arm.com \
    --cc=apopple@nvidia.com \
    --cc=baohua@kernel.org \
    --cc=baolin.wang@linux.alibaba.com \
    --cc=byungchul@sk.com \
    --cc=catalin.marinas@arm.com \
    --cc=cl@gentwo.org \
    --cc=corbet@lwn.net \
    --cc=dave.hansen@linux.intel.com \
    --cc=dev.jain@arm.com \
    --cc=gourry@gourry.net \
    --cc=hannes@cmpxchg.org \
    --cc=hughd@google.com \
    --cc=jack@suse.cz \
    --cc=jackmanb@google.com \
    --cc=jannh@google.com \
    --cc=jglisse@google.com \
    --cc=joshua.hahnjy@gmail.com \
    --cc=kas@kernel.org \
    --cc=lance.yang@linux.dev \
    --cc=liam@infradead.org \
    --cc=linux-doc@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=linux-trace-kernel@vger.kernel.org \
    --cc=ljs@kernel.org \
    --cc=mathieu.desnoyers@efficios.com \
    --cc=matthew.brost@intel.com \
    --cc=mhiramat@kernel.org \
    --cc=mhocko@suse.com \
    --cc=npache@redhat.com \
    --cc=peterx@redhat.com \
    --cc=pfalcato@suse.de \
    --cc=rakie.kim@sk.com \
    --cc=raquini@redhat.com \
    --cc=rdunlap@infradead.org \
    --cc=richard.weiyang@gmail.com \
    --cc=rientjes@google.com \
    --cc=rostedt@goodmis.org \
    --cc=rppt@kernel.org \
    --cc=ryan.roberts@arm.com \
    --cc=shivankg@amd.com \
    --cc=sunnanyong@huawei.com \
    --cc=surenb@google.com \
    --cc=thomas.hellstrom@linux.intel.com \
    --cc=tiwai@suse.de \
    --cc=usama.arif@linux.dev \
    --cc=usamaarif642@gmail.com \
    --cc=vbabka@suse.cz \
    --cc=vishal.moola@gmail.com \
    --cc=wangkefeng.wang@huawei.com \
    --cc=will@kernel.org \
    --cc=willy@infradead.org \
    --cc=yang@os.amperecomputing.com \
    --cc=ying.huang@linux.alibaba.com \
    --cc=ziy@nvidia.com \
    --cc=zokeefe@google.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